Extract metrics with opentelemetry framework #243
|
|
@ -11,6 +11,12 @@ dependencies = [
|
|||
"memchr",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "anyhow"
|
||||
version = "1.0.56"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "4361135be9122e0870de935d7c439aef945b9f9ddd4199a553b5270b49c82a27"
|
||||
|
||||
[[package]]
|
||||
name = "arc-swap"
|
||||
version = "1.5.0"
|
||||
|
|
@ -23,6 +29,27 @@ version = "0.5.2"
|
|||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "23b62fc65de8e4e7f52534fb52b0f3ed04746ae267519eef2a83941e8085068b"
|
||||
|
||||
[[package]]
|
||||
name = "async-stream"
|
||||
version = "0.3.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "dad5c83079eae9969be7fadefe640a1c566901f05ff91ab221de4b6f68d9507e"
|
||||
dependencies = [
|
||||
"async-stream-impl",
|
||||
"futures-core",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "async-stream-impl"
|
||||
version = "0.3.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "10f203db73a71dfa2fb6dd22763990fa26f3d2625a6da2da900d23b87d26be27"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "async-trait"
|
||||
version = "0.1.52"
|
||||
|
|
@ -369,6 +396,15 @@ dependencies = [
|
|||
"unicode-width",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "cloudabi"
|
||||
version = "0.0.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "ddfc5b9aa5d4507acaf872de71051dfd0e309860e88966e1051e462a077aac4f"
|
||||
dependencies = [
|
||||
"bitflags",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "core-foundation"
|
||||
version = "0.9.3"
|
||||
|
|
@ -403,6 +439,16 @@ dependencies = [
|
|||
"cfg-if",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "crossbeam-channel"
|
||||
version = "0.5.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "e54ea8bc3fb1ee042f5aace6e3c6e025d3874866da222930f70ce62aceba0bfa"
|
||||
dependencies = [
|
||||
"cfg-if",
|
||||
"crossbeam-utils",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "crossbeam-epoch"
|
||||
version = "0.9.7"
|
||||
|
|
@ -618,6 +664,12 @@ dependencies = [
|
|||
"instant",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "fixedbitset"
|
||||
version = "0.4.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "279fb028e20b3c4c320317955b77c5e0c9701f05a1d309905d6fc702cdc5053e"
|
||||
|
||||
[[package]]
|
||||
name = "fnv"
|
||||
version = "1.0.7"
|
||||
|
|
@ -659,6 +711,12 @@ dependencies = [
|
|||
"winapi",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "fuchsia-cprng"
|
||||
version = "0.1.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "a06f77d526c1a601b7c4cdd98f54b5eaabffc14d5f2f0296febdc7f357c6d3ba"
|
||||
|
||||
[[package]]
|
||||
name = "futures"
|
||||
version = "0.3.21"
|
||||
|
|
@ -759,28 +817,28 @@ dependencies = [
|
|||
|
||||
[[package]]
|
||||
name = "garage"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
dependencies = [
|
||||
"async-trait",
|
||||
"aws-sdk-s3",
|
||||
"bytes 1.1.0",
|
||||
"futures",
|
||||
"futures-util",
|
||||
"garage_admin",
|
||||
"garage_api",
|
||||
"garage_model 0.6.0",
|
||||
"garage_rpc 0.6.0",
|
||||
"garage_table 0.6.0",
|
||||
"garage_util 0.6.0",
|
||||
"garage_model 0.7.0",
|
||||
"garage_rpc 0.7.0",
|
||||
"garage_table 0.7.0",
|
||||
"garage_util 0.7.0",
|
||||
"garage_web",
|
||||
"git-version",
|
||||
"hex",
|
||||
"http",
|
||||
"hyper",
|
||||
"kuska-sodiumoxide",
|
||||
"log",
|
||||
"netapp",
|
||||
"netapp 0.4.0",
|
||||
"pretty_env_logger",
|
||||
"rand",
|
||||
"rand 0.8.5",
|
||||
"rmp-serde 0.15.5",
|
||||
"serde",
|
||||
"serde_bytes",
|
||||
|
|
@ -789,11 +847,29 @@ dependencies = [
|
|||
"structopt",
|
||||
"tokio",
|
||||
"toml",
|
||||
"tracing",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "garage_admin"
|
||||
version = "0.7.0"
|
||||
dependencies = [
|
||||
"futures",
|
||||
"futures-util",
|
||||
"garage_util 0.7.0",
|
||||
"hex",
|
||||
"http",
|
||||
"hyper",
|
||||
"opentelemetry",
|
||||
"opentelemetry-otlp",
|
||||
"opentelemetry-prometheus",
|
||||
"prometheus",
|
||||
"tracing",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "garage_api"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
dependencies = [
|
||||
"base64",
|
||||
"bytes 1.1.0",
|
||||
|
|
@ -803,9 +879,9 @@ dependencies = [
|
|||
"form_urlencoded",
|
||||
"futures",
|
||||
"futures-util",
|
||||
"garage_model 0.6.0",
|
||||
"garage_table 0.6.0",
|
||||
"garage_util 0.6.0",
|
||||
"garage_model 0.7.0",
|
||||
"garage_table 0.7.0",
|
||||
"garage_util 0.7.0",
|
||||
"hex",
|
||||
"hmac",
|
||||
"http",
|
||||
|
|
@ -813,10 +889,10 @@ dependencies = [
|
|||
"httpdate 0.3.2",
|
||||
"hyper",
|
||||
"idna",
|
||||
"log",
|
||||
"md-5",
|
||||
"multer",
|
||||
"nom",
|
||||
"opentelemetry",
|
||||
"percent-encoding",
|
||||
"pin-project 1.0.10",
|
||||
"quick-xml",
|
||||
|
|
@ -826,6 +902,7 @@ dependencies = [
|
|||
"serde_json",
|
||||
"sha2",
|
||||
"tokio",
|
||||
"tracing",
|
||||
"url",
|
||||
]
|
||||
|
||||
|
|
@ -844,8 +921,8 @@ dependencies = [
|
|||
"garage_util 0.5.1",
|
||||
"hex",
|
||||
"log",
|
||||
"netapp",
|
||||
"rand",
|
||||
"netapp 0.3.1",
|
||||
"rand 0.8.5",
|
||||
"rmp-serde 0.15.5",
|
||||
"serde",
|
||||
"serde_bytes",
|
||||
|
|
@ -856,7 +933,7 @@ dependencies = [
|
|||
|
||||
[[package]]
|
||||
name = "garage_model"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
dependencies = [
|
||||
"arc-swap",
|
||||
"async-trait",
|
||||
|
|
@ -864,18 +941,19 @@ dependencies = [
|
|||
"futures",
|
||||
"futures-util",
|
||||
"garage_model 0.5.1",
|
||||
"garage_rpc 0.6.0",
|
||||
"garage_table 0.6.0",
|
||||
"garage_util 0.6.0",
|
||||
"garage_rpc 0.7.0",
|
||||
"garage_table 0.7.0",
|
||||
"garage_util 0.7.0",
|
||||
"hex",
|
||||
"log",
|
||||
"netapp",
|
||||
"rand",
|
||||
"netapp 0.4.0",
|
||||
"opentelemetry",
|
||||
"rand 0.8.5",
|
||||
"rmp-serde 0.15.5",
|
||||
"serde",
|
||||
"serde_bytes",
|
||||
"sled",
|
||||
"tokio",
|
||||
"tracing",
|
||||
"zstd",
|
||||
]
|
||||
|
||||
|
|
@ -896,8 +974,8 @@ dependencies = [
|
|||
"hyper",
|
||||
"kuska-sodiumoxide",
|
||||
"log",
|
||||
"netapp",
|
||||
"rand",
|
||||
"netapp 0.3.1",
|
||||
"rand 0.8.5",
|
||||
"rmp-serde 0.15.5",
|
||||
"serde",
|
||||
"serde_bytes",
|
||||
|
|
@ -908,25 +986,26 @@ dependencies = [
|
|||
|
||||
[[package]]
|
||||
name = "garage_rpc"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
dependencies = [
|
||||
"arc-swap",
|
||||
"async-trait",
|
||||
"bytes 1.1.0",
|
||||
"futures",
|
||||
"futures-util",
|
||||
"garage_util 0.6.0",
|
||||
"garage_admin",
|
||||
"garage_util 0.7.0",
|
||||
"gethostname",
|
||||
"hex",
|
||||
"hyper",
|
||||
"k8s-openapi",
|
||||
"kube",
|
||||
"kuska-sodiumoxide",
|
||||
"log",
|
||||
"netapp",
|
||||
"netapp 0.4.0",
|
||||
"openssl",
|
||||
"opentelemetry",
|
||||
"pnet",
|
||||
"rand",
|
||||
"rand 0.8.5",
|
||||
"rmp-serde 0.15.5",
|
||||
"schemars",
|
||||
"serde",
|
||||
|
|
@ -934,6 +1013,7 @@ dependencies = [
|
|||
"serde_json",
|
||||
"tokio",
|
||||
"tokio-stream",
|
||||
"tracing",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
|
|
@ -950,7 +1030,7 @@ dependencies = [
|
|||
"garage_util 0.5.1",
|
||||
"hexdump",
|
||||
"log",
|
||||
"rand",
|
||||
"rand 0.8.5",
|
||||
"rmp-serde 0.15.5",
|
||||
"serde",
|
||||
"serde_bytes",
|
||||
|
|
@ -960,22 +1040,23 @@ dependencies = [
|
|||
|
||||
[[package]]
|
||||
name = "garage_table"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
dependencies = [
|
||||
"async-trait",
|
||||
"bytes 1.1.0",
|
||||
"futures",
|
||||
"futures-util",
|
||||
"garage_rpc 0.6.0",
|
||||
"garage_util 0.6.0",
|
||||
"garage_rpc 0.7.0",
|
||||
"garage_util 0.7.0",
|
||||
"hexdump",
|
||||
"log",
|
||||
"rand",
|
||||
"opentelemetry",
|
||||
"rand 0.8.5",
|
||||
"rmp-serde 0.15.5",
|
||||
"serde",
|
||||
"serde_bytes",
|
||||
"sled",
|
||||
"tokio",
|
||||
"tracing",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
|
|
@ -992,8 +1073,8 @@ dependencies = [
|
|||
"http",
|
||||
"hyper",
|
||||
"log",
|
||||
"netapp",
|
||||
"rand",
|
||||
"netapp 0.3.1",
|
||||
"rand 0.8.5",
|
||||
"rmp-serde 0.15.5",
|
||||
"serde",
|
||||
"serde_json",
|
||||
|
|
@ -1006,7 +1087,7 @@ dependencies = [
|
|||
|
||||
[[package]]
|
||||
name = "garage_util"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
dependencies = [
|
||||
"blake2",
|
||||
"chrono",
|
||||
|
|
@ -1017,9 +1098,9 @@ dependencies = [
|
|||
"hyper",
|
||||
"k8s-openapi",
|
||||
"kube",
|
||||
"log",
|
||||
"netapp",
|
||||
"rand",
|
||||
"netapp 0.4.0",
|
||||
"opentelemetry",
|
||||
"rand 0.8.5",
|
||||
"rmp-serde 0.15.5",
|
||||
"serde",
|
||||
"serde_json",
|
||||
|
|
@ -1027,23 +1108,25 @@ dependencies = [
|
|||
"sled",
|
||||
"tokio",
|
||||
"toml",
|
||||
"tracing",
|
||||
"xxhash-rust",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "garage_web"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
dependencies = [
|
||||
"err-derive 0.3.1",
|
||||
"futures",
|
||||
"garage_api",
|
||||
"garage_model 0.6.0",
|
||||
"garage_table 0.6.0",
|
||||
"garage_util 0.6.0",
|
||||
"garage_model 0.7.0",
|
||||
"garage_table 0.7.0",
|
||||
"garage_util 0.7.0",
|
||||
"http",
|
||||
"hyper",
|
||||
"log",
|
||||
"opentelemetry",
|
||||
"percent-encoding",
|
||||
"tracing",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
|
|
@ -1161,7 +1244,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
|
|||
checksum = "e40283dadb02f3af778878be1d717b17b4e4ab92e1d935ab03a730b0542905f2"
|
||||
dependencies = [
|
||||
"arrayvec",
|
||||
"itertools",
|
||||
"itertools 0.4.19",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
|
|
@ -1346,6 +1429,15 @@ version = "0.4.19"
|
|||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "c4a9b56eb56058f43dc66e58f40a214b2ccbc9f3df51861b63d51dec7b65bc3f"
|
||||
|
||||
[[package]]
|
||||
name = "itertools"
|
||||
version = "0.10.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "a9a9d19fa1e79b6215ff29b9d6880b706147f16e9b1dbb1e4e5947b5b02bc5e3"
|
||||
dependencies = [
|
||||
"either",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "itoa"
|
||||
version = "1.0.1"
|
||||
|
|
@ -1671,6 +1763,12 @@ dependencies = [
|
|||
"version_check",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "multimap"
|
||||
version = "0.8.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "e5ce46fe64a9d73be07dcbe690a38ce1b293be448fd8ce1e6c1b8062c9f72c6a"
|
||||
|
||||
[[package]]
|
||||
name = "native-tls"
|
||||
version = "0.2.8"
|
||||
|
|
@ -1711,6 +1809,32 @@ dependencies = [
|
|||
"tokio-util 0.6.9",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "netapp"
|
||||
version = "0.4.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "22c545a13b0c47b47e8052b35c4884dbe33c9ea62607371b0f4f1b0490cafd38"
|
||||
dependencies = [
|
||||
"arc-swap",
|
||||
"async-trait",
|
||||
"bytes 0.6.0",
|
||||
"cfg-if",
|
||||
"err-derive 0.2.4",
|
||||
"futures",
|
||||
"hex",
|
||||
"kuska-handshake",
|
||||
"kuska-sodiumoxide",
|
||||
"log",
|
||||
"opentelemetry",
|
||||
"opentelemetry-contrib",
|
||||
"rand 0.5.6",
|
||||
"rmp-serde 0.14.4",
|
||||
"serde",
|
||||
"tokio",
|
||||
"tokio-stream",
|
||||
"tokio-util 0.6.9",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "nom"
|
||||
version = "7.1.0"
|
||||
|
|
@ -1824,6 +1948,68 @@ dependencies = [
|
|||
"vcpkg",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "opentelemetry"
|
||||
version = "0.17.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "6105e89802af13fdf48c49d7646d3b533a70e536d818aae7e78ba0433d01acb8"
|
||||
dependencies = [
|
||||
"async-trait",
|
||||
"crossbeam-channel",
|
||||
"dashmap",
|
||||
"fnv",
|
||||
"futures-channel",
|
||||
"futures-executor",
|
||||
"futures-util",
|
||||
"js-sys",
|
||||
"lazy_static",
|
||||
"percent-encoding",
|
||||
"pin-project 1.0.10",
|
||||
"rand 0.8.5",
|
||||
"thiserror",
|
||||
"tokio",
|
||||
"tokio-stream",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "opentelemetry-contrib"
|
||||
version = "0.9.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "85637add8f60bb4cac673469c14f47a329c6cec7365c72d72cd32f2d104a721a"
|
||||
dependencies = [
|
||||
"lazy_static",
|
||||
"opentelemetry",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "opentelemetry-otlp"
|
||||
version = "0.10.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "9d1a6ca9de4c8b00aa7f1a153bd76cb263287155cec642680d79d98706f3d28a"
|
||||
dependencies = [
|
||||
"async-trait",
|
||||
"futures",
|
||||
"futures-util",
|
||||
"http",
|
||||
"opentelemetry",
|
||||
"prost",
|
||||
"thiserror",
|
||||
"tokio",
|
||||
"tonic",
|
||||
"tonic-build",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "opentelemetry-prometheus"
|
||||
version = "0.10.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "9328977e479cebe12ce0d3fcecdaea4721d234895a9440c5b5dfd113f0594ac6"
|
||||
dependencies = [
|
||||
"opentelemetry",
|
||||
"prometheus",
|
||||
"protobuf",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "ordered-float"
|
||||
version = "2.10.0"
|
||||
|
|
@ -1841,7 +2027,17 @@ checksum = "7d17b78036a60663b797adeaee46f5c9dfebb86948d1255007a1d6be0271ff99"
|
|||
dependencies = [
|
||||
"instant",
|
||||
"lock_api",
|
||||
"parking_lot_core",
|
||||
"parking_lot_core 0.8.5",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "parking_lot"
|
||||
version = "0.12.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "87f5ec2493a61ac0506c0f4199f99070cbe83857b0337006a30f3e6719b8ef58"
|
||||
dependencies = [
|
||||
"lock_api",
|
||||
"parking_lot_core 0.9.1",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
|
|
@ -1858,6 +2054,19 @@ dependencies = [
|
|||
"winapi",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "parking_lot_core"
|
||||
version = "0.9.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "28141e0cc4143da2443301914478dc976a61ffdb3f043058310c70df2fed8954"
|
||||
dependencies = [
|
||||
"cfg-if",
|
||||
"libc",
|
||||
"redox_syscall",
|
||||
"smallvec",
|
||||
"windows-sys",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pem"
|
||||
version = "0.8.3"
|
||||
|
|
@ -1875,6 +2084,16 @@ version = "2.1.0"
|
|||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "d4fd5641d01c8f18a23da7b6fe29298ff4b55afcccdf78973b24cf3175fee32e"
|
||||
|
||||
[[package]]
|
||||
name = "petgraph"
|
||||
version = "0.6.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "4a13a2fa9d0b63e5f22328828741e523766fff0ee9e779316902290dff3f824f"
|
||||
dependencies = [
|
||||
"fixedbitset",
|
||||
"indexmap",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pin-project"
|
||||
version = "0.4.29"
|
||||
|
|
@ -2076,6 +2295,80 @@ dependencies = [
|
|||
"unicode-xid",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "prometheus"
|
||||
version = "0.13.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "b7f64969ffd5dd8f39bd57a68ac53c163a095ed9d0fb707146da1b27025a3504"
|
||||
dependencies = [
|
||||
"cfg-if",
|
||||
"fnv",
|
||||
"lazy_static",
|
||||
"memchr",
|
||||
"parking_lot 0.11.2",
|
||||
"protobuf",
|
||||
"thiserror",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "prost"
|
||||
version = "0.9.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "444879275cb4fd84958b1a1d5420d15e6fcf7c235fe47f053c9c2a80aceb6001"
|
||||
dependencies = [
|
||||
"bytes 1.1.0",
|
||||
"prost-derive",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "prost-build"
|
||||
version = "0.9.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "62941722fb675d463659e49c4f3fe1fe792ff24fe5bbaa9c08cd3b98a1c354f5"
|
||||
dependencies = [
|
||||
"bytes 1.1.0",
|
||||
"heck",
|
||||
"itertools 0.10.3",
|
||||
"lazy_static",
|
||||
"log",
|
||||
"multimap",
|
||||
"petgraph",
|
||||
"prost",
|
||||
"prost-types",
|
||||
"regex",
|
||||
"tempfile",
|
||||
"which",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "prost-derive"
|
||||
version = "0.9.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "f9cc1a3263e07e0bf68e96268f37665207b49560d98739662cdfaae215c720fe"
|
||||
dependencies = [
|
||||
"anyhow",
|
||||
"itertools 0.10.3",
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "prost-types"
|
||||
version = "0.9.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "534b7a0e836e3c482d2693070f982e39e7611da9695d4d1f5a4b186b51faef0a"
|
||||
dependencies = [
|
||||
"bytes 1.1.0",
|
||||
"prost",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "protobuf"
|
||||
version = "2.27.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "cf7e6d18738ecd0902d30d1ad232c9125985a3422929b16c65517b38adc14f96"
|
||||
|
||||
[[package]]
|
||||
name = "quick-error"
|
||||
version = "1.2.3"
|
||||
|
|
@ -2101,6 +2394,19 @@ dependencies = [
|
|||
"proc-macro2",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "rand"
|
||||
version = "0.5.6"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "c618c47cd3ebd209790115ab837de41425723956ad3ce2e6a7f09890947cacb9"
|
||||
dependencies = [
|
||||
"cloudabi",
|
||||
"fuchsia-cprng",
|
||||
"libc",
|
||||
"rand_core 0.3.1",
|
||||
"winapi",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "rand"
|
||||
version = "0.8.5"
|
||||
|
|
@ -2109,7 +2415,7 @@ checksum = "34af8d1a0e25924bc5b7c43c079c942339d8f0a8b57c39049bef581b46327404"
|
|||
dependencies = [
|
||||
"libc",
|
||||
"rand_chacha",
|
||||
"rand_core",
|
||||
"rand_core 0.6.3",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
|
|
@ -2119,9 +2425,24 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
|
|||
checksum = "e6c10a63a0fa32252be49d21e7709d4d4baf8d231c2dbce1eaa8141b9b127d88"
|
||||
dependencies = [
|
||||
"ppv-lite86",
|
||||
"rand_core",
|
||||
"rand_core 0.6.3",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "rand_core"
|
||||
version = "0.3.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "7a6fdeb83b075e8266dcc8762c22776f6877a63111121f5f8c7411e5be7eed4b"
|
||||
dependencies = [
|
||||
"rand_core 0.4.2",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "rand_core"
|
||||
version = "0.4.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "9c33a3c44ca05fa6f1807d8e6743f3824e8509beca625669633be0acbdf509dc"
|
||||
|
||||
[[package]]
|
||||
name = "rand_core"
|
||||
version = "0.6.3"
|
||||
|
|
@ -2481,7 +2802,7 @@ dependencies = [
|
|||
"fxhash",
|
||||
"libc",
|
||||
"log",
|
||||
"parking_lot",
|
||||
"parking_lot 0.11.2",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
|
|
@ -2544,8 +2865,8 @@ dependencies = [
|
|||
"bitflags",
|
||||
"cfg_aliases",
|
||||
"libc",
|
||||
"parking_lot",
|
||||
"parking_lot_core",
|
||||
"parking_lot 0.11.2",
|
||||
"parking_lot_core 0.8.5",
|
||||
"static_init_macro",
|
||||
"winapi",
|
||||
]
|
||||
|
|
@ -2722,6 +3043,7 @@ dependencies = [
|
|||
"mio",
|
||||
"num_cpus",
|
||||
"once_cell",
|
||||
"parking_lot 0.12.0",
|
||||
"pin-project-lite",
|
||||
"signal-hook-registry",
|
||||
"socket2",
|
||||
|
|
@ -2821,6 +3143,49 @@ dependencies = [
|
|||
"serde",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "tonic"
|
||||
version = "0.6.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "ff08f4649d10a70ffa3522ca559031285d8e421d727ac85c60825761818f5d0a"
|
||||
dependencies = [
|
||||
"async-stream",
|
||||
"async-trait",
|
||||
"base64",
|
||||
"bytes 1.1.0",
|
||||
"futures-core",
|
||||
"futures-util",
|
||||
"h2",
|
||||
"http",
|
||||
"http-body",
|
||||
"hyper",
|
||||
"hyper-timeout",
|
||||
"percent-encoding",
|
||||
"pin-project 1.0.10",
|
||||
"prost",
|
||||
"prost-derive",
|
||||
"tokio",
|
||||
"tokio-stream",
|
||||
"tokio-util 0.6.9",
|
||||
"tower",
|
||||
"tower-layer",
|
||||
"tower-service",
|
||||
"tracing",
|
||||
"tracing-futures",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "tonic-build"
|
||||
version = "0.6.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "9403f1bafde247186684b230dc6f38b5cd514584e8bec1dd32514be4745fa757"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"prost-build",
|
||||
"quote",
|
||||
"syn",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "tower"
|
||||
version = "0.4.12"
|
||||
|
|
@ -2829,8 +3194,11 @@ checksum = "9a89fd63ad6adf737582df5db40d286574513c69a11dac5214dc3b5603d6713e"
|
|||
dependencies = [
|
||||
"futures-core",
|
||||
"futures-util",
|
||||
"indexmap",
|
||||
"pin-project 1.0.10",
|
||||
"pin-project-lite",
|
||||
"rand 0.8.5",
|
||||
"slab",
|
||||
"tokio",
|
||||
"tokio-util 0.7.0",
|
||||
"tower-layer",
|
||||
|
|
@ -2901,6 +3269,16 @@ dependencies = [
|
|||
"lazy_static",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "tracing-futures"
|
||||
version = "0.2.5"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "97d095ae15e245a057c8e8451bab9b3ee1e1f68e9ba2b4fbc18d0ac5237835f2"
|
||||
dependencies = [
|
||||
"pin-project 1.0.10",
|
||||
"tracing",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "treediff"
|
||||
version = "3.0.2"
|
||||
|
|
@ -3092,6 +3470,17 @@ dependencies = [
|
|||
"untrusted",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "which"
|
||||
version = "4.2.4"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "2a5a7e487e921cf220206864a94a89b6c6905bfc19f1057fa26a4cb360e5c1d2"
|
||||
dependencies = [
|
||||
"either",
|
||||
"lazy_static",
|
||||
"libc",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "winapi"
|
||||
version = "0.3.9"
|
||||
|
|
@ -3123,6 +3512,49 @@ version = "0.4.0"
|
|||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"
|
||||
|
||||
[[package]]
|
||||
name = "windows-sys"
|
||||
version = "0.32.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "3df6e476185f92a12c072be4a189a0210dcdcf512a1891d6dff9edb874deadc6"
|
||||
dependencies = [
|
||||
"windows_aarch64_msvc",
|
||||
"windows_i686_gnu",
|
||||
"windows_i686_msvc",
|
||||
"windows_x86_64_gnu",
|
||||
"windows_x86_64_msvc",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "windows_aarch64_msvc"
|
||||
version = "0.32.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "d8e92753b1c443191654ec532f14c199742964a061be25d77d7a96f09db20bf5"
|
||||
|
||||
[[package]]
|
||||
name = "windows_i686_gnu"
|
||||
version = "0.32.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "6a711c68811799e017b6038e0922cb27a5e2f43a2ddb609fe0b6f3eeda9de615"
|
||||
|
||||
[[package]]
|
||||
name = "windows_i686_msvc"
|
||||
version = "0.32.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "146c11bb1a02615db74680b32a68e2d61f553cc24c4eb5b4ca10311740e44172"
|
||||
|
||||
[[package]]
|
||||
name = "windows_x86_64_gnu"
|
||||
version = "0.32.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "c912b12f7454c6620635bbff3450962753834be2a594819bd5e945af18ec64bc"
|
||||
|
||||
[[package]]
|
||||
name = "windows_x86_64_msvc"
|
||||
version = "0.32.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "504a2476202769977a040c6364301a3f65d0cc9e3fb08600b2bda150a0488316"
|
||||
|
||||
[[package]]
|
||||
name = "xmlparser"
|
||||
version = "0.13.3"
|
||||
|
|
|
|||
|
|
@ -4,9 +4,10 @@ members = [
|
|||
"src/rpc",
|
||||
"src/table",
|
||||
"src/model",
|
||||
"src/admin",
|
||||
"src/api",
|
||||
"src/web",
|
||||
"src/garage",
|
||||
"src/garage"
|
||||
]
|
||||
|
||||
[profile.dev]
|
||||
|
|
|
|||
|
|
@ -44,6 +44,9 @@ root_domain = ".s3.garage.localhost"
|
|||
bind_addr = "0.0.0.0:$((3920+$count))"
|
||||
root_domain = ".web.garage.localhost"
|
||||
index = "index.html"
|
||||
|
||||
[admin]
|
||||
api_bind_addr = "0.0.0.0:$((9900+$count))"
|
||||
EOF
|
||||
|
||||
echo -en "$LABEL configuration written to $CONF_PATH\n"
|
||||
|
|
|
|||
|
|
@ -0,0 +1,3 @@
|
|||
COMPOSE_PROJECT_NAME=telemetry
|
||||
OTEL_COLLECT_TAG=0.44.0
|
||||
ELASTIC_BUNDLE_TAG=7.17.0
|
||||
|
|
@ -0,0 +1,10 @@
|
|||
apm-server:
|
||||
# Defines the host and port the server is listening on. Use "unix:/path/to.sock" to listen on a unix domain socket.
|
||||
host: "0.0.0.0:8200"
|
||||
#-------------------------- Elasticsearch output --------------------------
|
||||
output.elasticsearch:
|
||||
# Array of hosts to connect to.
|
||||
# Scheme and port can be left out and will be set to the default (`http` and `9200`).
|
||||
# In case you specify and additional path, the scheme is required: `http://localhost:9200/path`.
|
||||
# IPv6 addresses should always be defined as: `https://[2001:db8::1]:9200`.
|
||||
hosts: ["localhost:9200"]
|
||||
|
|
@ -0,0 +1,69 @@
|
|||
version: "2"
|
||||
services:
|
||||
|
||||
otel:
|
||||
image: otel/opentelemetry-collector-contrib:${OTEL_COLLECT_TAG}
|
||||
command: [ "--config=/etc/otel-config.yaml" ]
|
||||
volumes:
|
||||
- ./otel-config.yaml:/etc/otel-config.yaml
|
||||
network_mode: "host"
|
||||
|
||||
elastic:
|
||||
image: docker.elastic.co/elasticsearch/elasticsearch:${ELASTIC_BUNDLE_TAG}
|
||||
container_name: elastic
|
||||
environment:
|
||||
- "node.name=elastic"
|
||||
- "http.port=9200"
|
||||
- "cluster.name=es-docker-cluster"
|
||||
- "discovery.type=single-node"
|
||||
- "bootstrap.memory_lock=true"
|
||||
- "ES_JAVA_OPTS=-Xms512m -Xmx512m"
|
||||
ulimits:
|
||||
memlock:
|
||||
soft: -1
|
||||
hard: -1
|
||||
nofile: 65536
|
||||
volumes:
|
||||
- "es_data:/usr/share/elasticsearch/data"
|
||||
network_mode: "host"
|
||||
|
||||
# kibana instance and collectors
|
||||
# see https://www.elastic.co/guide/en/elastic-stack-get-started/current/get-started-docker.html
|
||||
kibana:
|
||||
image: docker.elastic.co/kibana/kibana:${ELASTIC_BUNDLE_TAG}
|
||||
container_name: kibana
|
||||
environment:
|
||||
SERVER_NAME: "kibana.local"
|
||||
# ELASTICSEARCH_URL: "http://localhost:9700"
|
||||
ELASTICSEARCH_HOSTS: "http://localhost:9200"
|
||||
depends_on: [ 'elastic' ]
|
||||
network_mode: "host"
|
||||
|
||||
apm:
|
||||
image: docker.elastic.co/apm/apm-server:${ELASTIC_BUNDLE_TAG}
|
||||
container_name: apm
|
||||
volumes:
|
||||
- "./apm-config.yaml:/usr/share/apm-server/apm-server.yml:ro"
|
||||
depends_on: [ 'elastic' ]
|
||||
network_mode: "host"
|
||||
|
||||
grafana:
|
||||
# see https://grafana.com/docs/grafana/latest/installation/docker/
|
||||
image: "grafana/grafana:8.3.5"
|
||||
container_name: grafana
|
||||
# restart: unless-stopped
|
||||
environment:
|
||||
- "GF_INSTALL_PLUGINS=grafana-clock-panel,grafana-simple-json-datasource,grafana-piechart-panel,grafana-worldmap-panel,grafana-polystat-panel"
|
||||
network_mode: "host"
|
||||
volumes:
|
||||
# chown 472:472 if needed
|
||||
- grafana:/var/lib/grafana
|
||||
- ./grafana/provisioning/:/etc/grafana/provisioning/
|
||||
|
||||
volumes:
|
||||
es_data:
|
||||
driver: local
|
||||
grafana:
|
||||
driver: local
|
||||
metricbeat:
|
||||
driver: local
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
apiVersion: 1
|
||||
|
||||
datasources:
|
||||
- name: DS_ELASTICSEARCH
|
||||
type: elasticsearch
|
||||
access: proxy
|
||||
url: http://elastic:9700
|
||||
password: ''
|
||||
user: ''
|
||||
database: metricbeat-*
|
||||
basicAuth: false
|
||||
isDefault: true
|
||||
jsonData:
|
||||
esVersion: 70
|
||||
logLevelField: ''
|
||||
logMessageField: ''
|
||||
maxConcurrentShardRequests: 5
|
||||
timeField: "@timestamp"
|
||||
timeInterval: 10s
|
||||
readOnly: false
|
||||
|
|
@ -0,0 +1,47 @@
|
|||
receivers:
|
||||
# Data sources: metrics, traces
|
||||
otlp:
|
||||
protocols:
|
||||
grpc:
|
||||
endpoint: ":4317"
|
||||
http:
|
||||
endpoint: ":55681"
|
||||
# Data sources: metrics
|
||||
prometheus:
|
||||
config:
|
||||
scrape_configs:
|
||||
- job_name: "garage"
|
||||
scrape_interval: 5s
|
||||
static_configs:
|
||||
- targets: ["localhost:3909"]
|
||||
|
||||
exporters:
|
||||
logging:
|
||||
logLevel: info
|
||||
# see https://www.elastic.co/guide/en/apm/get-started/current/open-telemetry-elastic.html#open-telemetry-collector
|
||||
otlp/elastic:
|
||||
endpoint: "localhost:8200"
|
||||
tls:
|
||||
insecure: true
|
||||
|
||||
processors:
|
||||
batch:
|
||||
|
||||
extensions:
|
||||
health_check:
|
||||
pprof:
|
||||
endpoint: :1888
|
||||
zpages:
|
||||
endpoint: :55679
|
||||
|
||||
service:
|
||||
extensions: [pprof, zpages, health_check]
|
||||
pipelines:
|
||||
traces:
|
||||
receivers: [otlp]
|
||||
processors: [batch]
|
||||
exporters: [logging, otlp/elastic]
|
||||
metrics:
|
||||
receivers: [otlp, prometheus]
|
||||
processors: [batch]
|
||||
exporters: [logging, otlp/elastic]
|
||||
File diff suppressed because it is too large
Load Diff
|
|
@ -78,6 +78,7 @@ function refresh_toolchain {
|
|||
pkgs.clippy
|
||||
pkgs.rustfmt
|
||||
pkgs.perl
|
||||
pkgs.protobuf
|
||||
cargo2nix.packages.x86_64-linux.cargo2nix
|
||||
] else [])
|
||||
++
|
||||
|
|
|
|||
|
|
@ -0,0 +1,29 @@
|
|||
[package]
|
||||
name = "garage_admin"
|
||||
version = "0.7.0"
|
||||
authors = ["Maximilien Richer <code@mricher.fr>"]
|
||||
edition = "2018"
|
||||
license = "AGPL-3.0"
|
||||
description = "Administration and metrics REST HTTP server for Garage"
|
||||
repository = "https://git.deuxfleurs.fr/Deuxfleurs/garage"
|
||||
|
||||
[lib]
|
||||
path = "lib.rs"
|
||||
|
||||
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
|
||||
|
||||
[dependencies]
|
||||
garage_util = { version = "0.7.0", path = "../util" }
|
||||
|
||||
hex = "0.4"
|
||||
|
||||
futures = "0.3"
|
||||
futures-util = "0.3"
|
||||
http = "0.2"
|
||||
hyper = "0.14"
|
||||
tracing = "0.1.30"
|
||||
|
||||
opentelemetry = { version = "0.17", features = [ "rt-tokio" ] }
|
||||
opentelemetry-prometheus = "0.10"
|
||||
opentelemetry-otlp = "0.10"
|
||||
prometheus = "0.13"
|
||||
|
|
@ -0,0 +1,6 @@
|
|||
//! Crate for handling the admin and metric HTTP APIs
|
||||
#[macro_use]
|
||||
extern crate tracing;
|
||||
|
||||
pub mod metrics;
|
||||
pub mod tracing_setup;
|
||||
|
|
@ -0,0 +1,146 @@
|
|||
use std::convert::Infallible;
|
||||
use std::net::SocketAddr;
|
||||
use std::sync::Arc;
|
||||
use std::time::SystemTime;
|
||||
|
||||
use futures::future::*;
|
||||
use hyper::{
|
||||
header::CONTENT_TYPE,
|
||||
service::{make_service_fn, service_fn},
|
||||
Body, Method, Request, Response, Server,
|
||||
};
|
||||
|
||||
use opentelemetry::{
|
||||
global,
|
||||
metrics::{BoundCounter, BoundValueRecorder},
|
||||
trace::{FutureExt, TraceContextExt, Tracer},
|
||||
Context,
|
||||
};
|
||||
use opentelemetry_prometheus::PrometheusExporter;
|
||||
|
||||
use prometheus::{Encoder, TextEncoder};
|
||||
|
||||
use garage_util::error::Error as GarageError;
|
||||
use garage_util::metrics::*;
|
||||
|
||||
// serve_req on metric endpoint
|
||||
async fn serve_req(
|
||||
req: Request<Body>,
|
||||
admin_server: Arc<AdminServer>,
|
||||
) -> Result<Response<Body>, hyper::Error> {
|
||||
info!("Receiving request at path {}", req.uri());
|
||||
let request_start = SystemTime::now();
|
||||
|
||||
admin_server.metrics.http_counter.add(1);
|
||||
|
||||
let response = match (req.method(), req.uri().path()) {
|
||||
(&Method::GET, "/metrics") => {
|
||||
let mut buffer = vec![];
|
||||
let encoder = TextEncoder::new();
|
||||
|
||||
let tracer = opentelemetry::global::tracer("garage");
|
||||
let metric_families = tracer.in_span("admin/gather_metrics", |_| {
|
||||
admin_server.exporter.registry().gather()
|
||||
});
|
||||
|
||||
encoder.encode(&metric_families, &mut buffer).unwrap();
|
||||
admin_server
|
||||
.metrics
|
||||
.http_body_gauge
|
||||
.record(buffer.len() as u64);
|
||||
|
||||
Response::builder()
|
||||
.status(200)
|
||||
.header(CONTENT_TYPE, encoder.format_type())
|
||||
.body(Body::from(buffer))
|
||||
.unwrap()
|
||||
}
|
||||
_ => Response::builder()
|
||||
.status(404)
|
||||
.body(Body::from("Not implemented"))
|
||||
.unwrap(),
|
||||
};
|
||||
|
||||
admin_server
|
||||
.metrics
|
||||
.http_req_histogram
|
||||
.record(request_start.elapsed().map_or(0.0, |d| d.as_secs_f64()));
|
||||
Ok(response)
|
||||
}
|
||||
|
||||
// AdminServer hold the admin server internal admin_server and the metric exporter
|
||||
pub struct AdminServer {
|
||||
exporter: PrometheusExporter,
|
||||
metrics: AdminServerMetrics,
|
||||
}
|
||||
|
||||
// GarageMetricadmin_server holds the metrics counter definition for Garage
|
||||
// FIXME: we would rather have that split up among the different libraries?
|
||||
struct AdminServerMetrics {
|
||||
http_counter: BoundCounter<u64>,
|
||||
http_body_gauge: BoundValueRecorder<u64>,
|
||||
http_req_histogram: BoundValueRecorder<f64>,
|
||||
}
|
||||
|
||||
impl AdminServer {
|
||||
/// init initilialize the AdminServer and background metric server
|
||||
pub fn init() -> AdminServer {
|
||||
let exporter = opentelemetry_prometheus::exporter().init();
|
||||
let meter = global::meter("garage/admin_server");
|
||||
AdminServer {
|
||||
exporter,
|
||||
metrics: AdminServerMetrics {
|
||||
http_counter: meter
|
||||
.u64_counter("admin.http_requests_total")
|
||||
.with_description("Total number of HTTP requests made.")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
http_body_gauge: meter
|
||||
.u64_value_recorder("admin.http_response_size_bytes")
|
||||
.with_description("The metrics HTTP response sizes in bytes.")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
http_req_histogram: meter
|
||||
.f64_value_recorder("admin.http_request_duration_seconds")
|
||||
.with_description("The HTTP request latencies in seconds.")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
},
|
||||
}
|
||||
}
|
||||
/// run execute the admin server on the designated HTTP port and listen for requests
|
||||
pub async fn run(
|
||||
self,
|
||||
bind_addr: SocketAddr,
|
||||
shutdown_signal: impl Future<Output = ()>,
|
||||
) -> Result<(), GarageError> {
|
||||
let admin_server = Arc::new(self);
|
||||
// For every connection, we must make a `Service` to handle all
|
||||
// incoming HTTP requests on said connection.
|
||||
let make_svc = make_service_fn(move |_conn| {
|
||||
let admin_server = admin_server.clone();
|
||||
// This is the `Service` that will handle the connection.
|
||||
// `service_fn` is a helper to convert a function that
|
||||
// returns a Response into a `Service`.
|
||||
async move {
|
||||
Ok::<_, Infallible>(service_fn(move |req| {
|
||||
let tracer = opentelemetry::global::tracer("garage");
|
||||
let span = tracer
|
||||
.span_builder("admin/request")
|
||||
.with_trace_id(gen_trace_id())
|
||||
.start(&tracer);
|
||||
|
||||
serve_req(req, admin_server.clone())
|
||||
.with_context(Context::current_with_span(span))
|
||||
}))
|
||||
}
|
||||
});
|
||||
|
||||
let server = Server::bind(&bind_addr).serve(make_svc);
|
||||
let graceful = server.with_graceful_shutdown(shutdown_signal);
|
||||
info!("Admin server listening on http://{}", bind_addr);
|
||||
|
||||
graceful.await?;
|
||||
Ok(())
|
||||
}
|
||||
}
|
||||
|
|
@ -0,0 +1,37 @@
|
|||
use std::time::Duration;
|
||||
|
||||
use opentelemetry::sdk::{
|
||||
trace::{self, IdGenerator, Sampler},
|
||||
Resource,
|
||||
};
|
||||
use opentelemetry::KeyValue;
|
||||
use opentelemetry_otlp::WithExportConfig;
|
||||
|
||||
use garage_util::data::*;
|
||||
use garage_util::error::*;
|
||||
|
||||
pub fn init_tracing(export_to: &str, node_id: Uuid) -> Result<(), Error> {
|
||||
let node_id = hex::encode(&node_id.as_slice()[..8]);
|
||||
|
||||
opentelemetry_otlp::new_pipeline()
|
||||
.tracing()
|
||||
.with_exporter(
|
||||
opentelemetry_otlp::new_exporter()
|
||||
.tonic()
|
||||
.with_endpoint(export_to)
|
||||
.with_timeout(Duration::from_secs(3)),
|
||||
)
|
||||
.with_trace_config(
|
||||
trace::config()
|
||||
.with_id_generator(IdGenerator::default())
|
||||
.with_sampler(Sampler::AlwaysOn)
|
||||
.with_resource(Resource::new(vec![
|
||||
KeyValue::new("service.name", "garage"),
|
||||
KeyValue::new("service.instance.id", node_id),
|
||||
])),
|
||||
)
|
||||
.install_batch(opentelemetry::runtime::Tokio)
|
||||
.ok_or_message("Unable to initialize tracing")?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
|
@ -1,6 +1,6 @@
|
|||
[package]
|
||||
name = "garage_api"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
authors = ["Alex Auvolat <alex@adnab.me>"]
|
||||
edition = "2018"
|
||||
license = "AGPL-3.0"
|
||||
|
|
@ -14,9 +14,9 @@ path = "lib.rs"
|
|||
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
|
||||
|
||||
[dependencies]
|
||||
garage_model = { version = "0.6.0", path = "../model" }
|
||||
garage_table = { version = "0.6.0", path = "../table" }
|
||||
garage_util = { version = "0.6.0", path = "../util" }
|
||||
garage_model = { version = "0.7.0", path = "../model" }
|
||||
garage_table = { version = "0.7.0", path = "../table" }
|
||||
garage_util = { version = "0.7.0", path = "../util" }
|
||||
|
||||
base64 = "0.13"
|
||||
bytes = "1.0"
|
||||
|
|
@ -26,7 +26,7 @@ err-derive = "0.3"
|
|||
hex = "0.4"
|
||||
hmac = "0.10"
|
||||
idna = "0.2"
|
||||
log = "0.4"
|
||||
tracing = "0.1.30"
|
||||
md-5 = "0.9"
|
||||
nom = "7.1"
|
||||
sha2 = "0.9"
|
||||
|
|
@ -49,3 +49,5 @@ serde_bytes = "0.11"
|
|||
serde_json = "1.0"
|
||||
quick-xml = { version = "0.21", features = [ "serialize" ] }
|
||||
url = "2.1"
|
||||
|
||||
opentelemetry = "0.17"
|
||||
|
|
|
|||
|
|
@ -7,8 +7,16 @@ use hyper::server::conn::AddrStream;
|
|||
use hyper::service::{make_service_fn, service_fn};
|
||||
use hyper::{Body, Method, Request, Response, Server};
|
||||
|
||||
use opentelemetry::{
|
||||
global,
|
||||
metrics::{Counter, ValueRecorder},
|
||||
trace::{FutureExt, TraceContextExt, Tracer},
|
||||
Context, KeyValue,
|
||||
};
|
||||
|
||||
use garage_util::data::*;
|
||||
use garage_util::error::Error as GarageError;
|
||||
use garage_util::metrics::{gen_trace_id, RecordDuration};
|
||||
|
||||
use garage_model::garage::Garage;
|
||||
use garage_model::key_table::Key;
|
||||
|
|
@ -30,6 +38,34 @@ use crate::s3_put::*;
|
|||
use crate::s3_router::{Authorization, Endpoint};
|
||||
use crate::s3_website::*;
|
||||
|
||||
struct ApiMetrics {
|
||||
request_counter: Counter<u64>,
|
||||
error_counter: Counter<u64>,
|
||||
request_duration: ValueRecorder<f64>,
|
||||
}
|
||||
|
||||
impl ApiMetrics {
|
||||
fn new() -> Self {
|
||||
let meter = global::meter("garage/api");
|
||||
Self {
|
||||
request_counter: meter
|
||||
.u64_counter("api.request_counter")
|
||||
.with_description("Number of API calls to the various S3 API endpoints")
|
||||
.init(),
|
||||
error_counter: meter
|
||||
.u64_counter("api.error_counter")
|
||||
.with_description(
|
||||
"Number of API calls to the various S3 API endpoints that resulted in errors",
|
||||
)
|
||||
.init(),
|
||||
request_duration: meter
|
||||
.f64_value_recorder("api.request_duration")
|
||||
.with_description("Duration of API calls to the various S3 API endpoints")
|
||||
.init(),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// Run the S3 API server
|
||||
pub async fn run_api_server(
|
||||
garage: Arc<Garage>,
|
||||
|
|
@ -37,13 +73,19 @@ pub async fn run_api_server(
|
|||
) -> Result<(), GarageError> {
|
||||
let addr = &garage.config.s3_api.api_bind_addr;
|
||||
|
||||
let metrics = Arc::new(ApiMetrics::new());
|
||||
|
||||
let service = make_service_fn(|conn: &AddrStream| {
|
||||
let garage = garage.clone();
|
||||
let metrics = metrics.clone();
|
||||
|
||||
let client_addr = conn.remote_addr();
|
||||
async move {
|
||||
Ok::<_, GarageError>(service_fn(move |req: Request<Body>| {
|
||||
let garage = garage.clone();
|
||||
handler(garage, req, client_addr)
|
||||
let metrics = metrics.clone();
|
||||
|
||||
handler(garage, metrics, req, client_addr)
|
||||
}))
|
||||
}
|
||||
});
|
||||
|
|
@ -59,13 +101,29 @@ pub async fn run_api_server(
|
|||
|
||||
async fn handler(
|
||||
garage: Arc<Garage>,
|
||||
metrics: Arc<ApiMetrics>,
|
||||
req: Request<Body>,
|
||||
addr: SocketAddr,
|
||||
) -> Result<Response<Body>, GarageError> {
|
||||
let uri = req.uri().clone();
|
||||
info!("{} {} {}", addr, req.method(), uri);
|
||||
debug!("{:?}", req);
|
||||
match handler_inner(garage.clone(), req).await {
|
||||
|
||||
let tracer = opentelemetry::global::tracer("garage");
|
||||
let span = tracer
|
||||
.span_builder("S3 API call (unknown)")
|
||||
.with_trace_id(gen_trace_id())
|
||||
.with_attributes(vec![
|
||||
KeyValue::new("method", format!("{}", req.method())),
|
||||
KeyValue::new("uri", req.uri().to_string()),
|
||||
])
|
||||
.start(&tracer);
|
||||
|
||||
let res = handler_stage2(garage.clone(), metrics, req)
|
||||
.with_context(Context::current_with_span(span))
|
||||
.await;
|
||||
|
||||
match res {
|
||||
Ok(x) => {
|
||||
debug!("{} {:?}", x.status(), x.headers());
|
||||
Ok(x)
|
||||
|
|
@ -92,7 +150,11 @@ async fn handler(
|
|||
}
|
||||
}
|
||||
|
||||
async fn handler_inner(garage: Arc<Garage>, req: Request<Body>) -> Result<Response<Body>, Error> {
|
||||
async fn handler_stage2(
|
||||
garage: Arc<Garage>,
|
||||
metrics: Arc<ApiMetrics>,
|
||||
req: Request<Body>,
|
||||
) -> Result<Response<Body>, Error> {
|
||||
let authority = req
|
||||
.headers()
|
||||
.get(header::HOST)
|
||||
|
|
@ -111,6 +173,46 @@ async fn handler_inner(garage: Arc<Garage>, req: Request<Body>) -> Result<Respon
|
|||
let (endpoint, bucket_name) = Endpoint::from_request(&req, bucket_name.map(ToOwned::to_owned))?;
|
||||
debug!("Endpoint: {:?}", endpoint);
|
||||
|
||||
let current_context = Context::current();
|
||||
let current_span = current_context.span();
|
||||
current_span.update_name::<String>(format!("S3 API {}", endpoint.name()));
|
||||
current_span.set_attribute(KeyValue::new("endpoint", endpoint.name()));
|
||||
current_span.set_attribute(KeyValue::new(
|
||||
"bucket",
|
||||
bucket_name.clone().unwrap_or_default(),
|
||||
));
|
||||
|
||||
let metrics_tags = &[KeyValue::new("api_endpoint", endpoint.name())];
|
||||
|
||||
let res = handler_stage3(garage, req, endpoint, bucket_name)
|
||||
.record_duration(&metrics.request_duration, &metrics_tags[..])
|
||||
.await;
|
||||
|
||||
metrics.request_counter.add(1, &metrics_tags[..]);
|
||||
|
||||
let status_code = match &res {
|
||||
Ok(r) => r.status(),
|
||||
Err(e) => e.http_status_code(),
|
||||
};
|
||||
if status_code.is_client_error() || status_code.is_server_error() {
|
||||
metrics.error_counter.add(
|
||||
1,
|
||||
&[
|
||||
metrics_tags[0].clone(),
|
||||
KeyValue::new("status_code", status_code.as_str().to_string()),
|
||||
],
|
||||
);
|
||||
}
|
||||
|
||||
res
|
||||
}
|
||||
|
||||
async fn handler_stage3(
|
||||
garage: Arc<Garage>,
|
||||
req: Request<Body>,
|
||||
endpoint: Endpoint,
|
||||
bucket_name: Option<String>,
|
||||
) -> Result<Response<Body>, Error> {
|
||||
// Some endpoints are processed early, before we even check for an API key
|
||||
if let Endpoint::PostObject = endpoint {
|
||||
return handle_post_object(garage, req, bucket_name.unwrap()).await;
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
//! Crate for serving a S3 compatible API
|
||||
#[macro_use]
|
||||
extern crate log;
|
||||
extern crate tracing;
|
||||
|
||||
pub mod error;
|
||||
pub use error::Error;
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
[package]
|
||||
name = "garage"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
authors = ["Alex Auvolat <alex@adnab.me>"]
|
||||
edition = "2018"
|
||||
license = "AGPL-3.0"
|
||||
|
|
@ -21,17 +21,18 @@ path = "tests/lib.rs"
|
|||
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
|
||||
|
||||
[dependencies]
|
||||
garage_api = { version = "0.6.0", path = "../api" }
|
||||
garage_model = { version = "0.6.0", path = "../model" }
|
||||
garage_rpc = { version = "0.6.0", path = "../rpc" }
|
||||
garage_table = { version = "0.6.0", path = "../table" }
|
||||
garage_util = { version = "0.6.0", path = "../util" }
|
||||
garage_web = { version = "0.6.0", path = "../web" }
|
||||
garage_api = { version = "0.7.0", path = "../api" }
|
||||
garage_model = { version = "0.7.0", path = "../model" }
|
||||
garage_rpc = { version = "0.7.0", path = "../rpc" }
|
||||
garage_table = { version = "0.7.0", path = "../table" }
|
||||
garage_util = { version = "0.7.0", path = "../util" }
|
||||
garage_web = { version = "0.7.0", path = "../web" }
|
||||
garage_admin = { version = "0.7.0", path = "../admin" }
|
||||
|
||||
bytes = "1.0"
|
||||
git-version = "0.3.4"
|
||||
hex = "0.4"
|
||||
log = "0.4"
|
||||
tracing = { version = "0.1.30", features = ["log-always"] }
|
||||
pretty_env_logger = "0.4"
|
||||
rand = "0.8"
|
||||
async-trait = "0.1.7"
|
||||
|
|
@ -49,8 +50,9 @@ futures = "0.3"
|
|||
futures-util = "0.3"
|
||||
tokio = { version = "1.0", default-features = false, features = ["rt", "rt-multi-thread", "io-util", "net", "time", "macros", "sync", "signal", "fs"] }
|
||||
|
||||
#netapp = { version = "0.3.0", git = "https://git.deuxfleurs.fr/lx/netapp" }
|
||||
netapp = "0.3.0"
|
||||
#netapp = { version = "0.3.2", git = "https://git.deuxfleurs.fr/lx/netapp" }
|
||||
#netapp = { version = "0.4", path = "../../../netapp" }
|
||||
netapp = "0.4"
|
||||
|
||||
[dev-dependencies]
|
||||
aws-sdk-s3 = "0.6"
|
||||
|
|
|
|||
|
|
@ -1,7 +1,5 @@
|
|||
use std::path::PathBuf;
|
||||
|
||||
use log::warn;
|
||||
|
||||
use garage_util::error::*;
|
||||
|
||||
pub const READ_KEY_ERROR: &str = "Unable to read node key. It will be generated by your garage node the first time is it launched. Ensure that your garage node is currently running. (The node key is supposed to be stored in your metadata directory.)";
|
||||
|
|
|
|||
|
|
@ -2,7 +2,7 @@
|
|||
//! Garage CLI, used to interact with a running Garage instance, and to launch a Garage instance
|
||||
|
||||
#[macro_use]
|
||||
extern crate log;
|
||||
extern crate tracing;
|
||||
|
||||
mod admin;
|
||||
mod cli;
|
||||
|
|
@ -55,7 +55,7 @@ struct Opt {
|
|||
#[tokio::main]
|
||||
async fn main() {
|
||||
if std::env::var("RUST_LOG").is_err() {
|
||||
std::env::set_var("RUST_LOG", "garage=info")
|
||||
std::env::set_var("RUST_LOG", "netapp=info,garage=info")
|
||||
}
|
||||
pretty_env_logger::init();
|
||||
sodiumoxide::init().expect("Unable to init sodiumoxide");
|
||||
|
|
@ -106,7 +106,7 @@ async fn cli_command(opt: Opt) -> Result<(), Error> {
|
|||
// Generate a temporary keypair for our RPC client
|
||||
let (_pk, sk) = sodiumoxide::crypto::sign::ed25519::gen_keypair();
|
||||
|
||||
let netapp = NetApp::new(network_key, sk);
|
||||
let netapp = NetApp::new(GARAGE_VERSION_TAG, network_key, sk);
|
||||
|
||||
// Find and parse the address of the target host
|
||||
let (id, addr) = if let Some(h) = opt.rpc_host {
|
||||
|
|
|
|||
|
|
@ -6,6 +6,8 @@ use garage_util::background::*;
|
|||
use garage_util::config::*;
|
||||
use garage_util::error::Error;
|
||||
|
||||
use garage_admin::metrics::*;
|
||||
use garage_admin::tracing_setup::*;
|
||||
use garage_api::run_api_server;
|
||||
use garage_model::garage::Garage;
|
||||
use garage_web::run_web_server;
|
||||
|
|
@ -34,6 +36,9 @@ pub async fn run_server(config_file: PathBuf) -> Result<(), Error> {
|
|||
.open()
|
||||
.expect("Unable to open sled DB");
|
||||
|
||||
info!("Initialize admin web server and metric backend...");
|
||||
let admin_server_init = AdminServer::init();
|
||||
|
||||
info!("Initializing background runner...");
|
||||
let watch_cancel = netapp::util::watch_ctrl_c();
|
||||
let (background, await_background_done) = BackgroundRunner::new(16, watch_cancel.clone());
|
||||
|
|
@ -41,9 +46,14 @@ pub async fn run_server(config_file: PathBuf) -> Result<(), Error> {
|
|||
info!("Initializing Garage main data store...");
|
||||
let garage = Garage::new(config.clone(), db, background);
|
||||
|
||||
info!("Initialize tracing...");
|
||||
if let Some(export_to) = config.admin.trace_sink {
|
||||
init_tracing(&export_to, garage.system.id)?;
|
||||
}
|
||||
|
||||
let run_system = tokio::spawn(garage.system.clone().run(watch_cancel.clone()));
|
||||
|
||||
info!("Crate admin RPC handler...");
|
||||
info!("Create admin RPC handler...");
|
||||
AdminRpcHandler::new(garage.clone());
|
||||
|
||||
info!("Initializing API server...");
|
||||
|
|
@ -58,6 +68,15 @@ pub async fn run_server(config_file: PathBuf) -> Result<(), Error> {
|
|||
wait_from(watch_cancel.clone()),
|
||||
));
|
||||
|
||||
let admin_server = if let Some(admin_bind_addr) = config.admin.api_bind_addr {
|
||||
info!("Configure and run admin web server...");
|
||||
Some(tokio::spawn(
|
||||
admin_server_init.run(admin_bind_addr, wait_from(watch_cancel.clone())),
|
||||
))
|
||||
} else {
|
||||
None
|
||||
};
|
||||
|
||||
// Stuff runs
|
||||
|
||||
// When a cancel signal is sent, stuff stops
|
||||
|
|
@ -67,6 +86,11 @@ pub async fn run_server(config_file: PathBuf) -> Result<(), Error> {
|
|||
if let Err(e) = web_server.await? {
|
||||
warn!("Web server exited with error: {}", e);
|
||||
}
|
||||
if let Some(a) = admin_server {
|
||||
if let Err(e) = a.await? {
|
||||
warn!("Admin web server exited with error: {}", e);
|
||||
}
|
||||
}
|
||||
|
||||
// Remove RPC handlers for system to break reference cycles
|
||||
garage.system.netapp.drop_all_handlers();
|
||||
|
|
|
|||
|
|
@ -65,6 +65,9 @@ root_domain = ".s3.garage"
|
|||
bind_addr = "127.0.0.1:{web_port}"
|
||||
root_domain = ".web.garage"
|
||||
index = "index.html"
|
||||
|
||||
[admin]
|
||||
api_bind_addr = "127.0.0.1:{admin_port}"
|
||||
"#,
|
||||
path = path.display(),
|
||||
secret = GARAGE_TEST_SECRET,
|
||||
|
|
@ -72,6 +75,7 @@ index = "index.html"
|
|||
api_port = port,
|
||||
rpc_port = port + 1,
|
||||
web_port = port + 2,
|
||||
admin_port = port + 3,
|
||||
);
|
||||
fs::write(path.join("config.toml"), config).expect("Could not write garage config file");
|
||||
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
[package]
|
||||
name = "garage_model"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
authors = ["Alex Auvolat <alex@adnab.me>"]
|
||||
edition = "2018"
|
||||
license = "AGPL-3.0"
|
||||
|
|
@ -14,16 +14,16 @@ path = "lib.rs"
|
|||
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
|
||||
|
||||
[dependencies]
|
||||
garage_rpc = { version = "0.6.0", path = "../rpc" }
|
||||
garage_table = { version = "0.6.0", path = "../table" }
|
||||
garage_util = { version = "0.6.0", path = "../util" }
|
||||
garage_rpc = { version = "0.7.0", path = "../rpc" }
|
||||
garage_table = { version = "0.7.0", path = "../table" }
|
||||
garage_util = { version = "0.7.0", path = "../util" }
|
||||
garage_model_050 = { package = "garage_model", version = "0.5.1" }
|
||||
|
||||
async-trait = "0.1.7"
|
||||
arc-swap = "1.0"
|
||||
err-derive = "0.3"
|
||||
hex = "0.4"
|
||||
log = "0.4"
|
||||
tracing = "0.1.30"
|
||||
rand = "0.8"
|
||||
zstd = { version = "0.9", default-features = false }
|
||||
|
||||
|
|
@ -36,6 +36,8 @@ serde_bytes = "0.11"
|
|||
futures = "0.3"
|
||||
futures-util = "0.3"
|
||||
tokio = { version = "1.0", default-features = false, features = ["rt", "rt-multi-thread", "io-util", "net", "time", "macros", "sync", "signal", "fs"] }
|
||||
opentelemetry = "0.17"
|
||||
|
||||
#netapp = { version = "0.3.0", git = "https://git.deuxfleurs.fr/lx/netapp" }
|
||||
netapp = "0.3.0"
|
||||
#netapp = { version = "0.4", path = "../../../netapp" }
|
||||
netapp = "0.4"
|
||||
|
|
|
|||
|
|
@ -5,16 +5,24 @@ use std::time::Duration;
|
|||
|
||||
use arc_swap::ArcSwapOption;
|
||||
use async_trait::async_trait;
|
||||
use serde::{Deserialize, Serialize};
|
||||
use zstd::stream::{decode_all as zstd_decode, Encoder};
|
||||
|
||||
use futures::future::*;
|
||||
use futures::select;
|
||||
use serde::{Deserialize, Serialize};
|
||||
use tokio::fs;
|
||||
use tokio::io::{AsyncReadExt, AsyncWriteExt};
|
||||
use tokio::sync::{watch, Mutex, Notify};
|
||||
use zstd::stream::{decode_all as zstd_decode, Encoder};
|
||||
|
||||
use opentelemetry::{
|
||||
trace::{FutureExt as OtelFutureExt, TraceContextExt, Tracer},
|
||||
Context, KeyValue,
|
||||
};
|
||||
|
||||
use garage_util::data::*;
|
||||
use garage_util::error::*;
|
||||
use garage_util::metrics::RecordDuration;
|
||||
use garage_util::sled_counter::SledCountedTree;
|
||||
use garage_util::time::*;
|
||||
use garage_util::tranquilizer::Tranquilizer;
|
||||
|
||||
|
|
@ -23,8 +31,8 @@ use garage_rpc::*;
|
|||
|
||||
use garage_table::replication::{TableReplication, TableShardedReplication};
|
||||
|
||||
use crate::block_metrics::*;
|
||||
use crate::block_ref_table::*;
|
||||
|
||||
use crate::garage::Garage;
|
||||
|
||||
/// Size under which data will be stored inlined in database instead of as files
|
||||
|
|
@ -148,12 +156,14 @@ pub struct BlockManager {
|
|||
|
||||
rc: sled::Tree,
|
||||
|
||||
resync_queue: sled::Tree,
|
||||
resync_queue: SledCountedTree,
|
||||
resync_notify: Notify,
|
||||
|
||||
system: Arc<System>,
|
||||
endpoint: Arc<Endpoint<BlockRpc, Self>>,
|
||||
pub(crate) garage: ArcSwapOption<Garage>,
|
||||
|
||||
metrics: BlockManagerMetrics,
|
||||
}
|
||||
|
||||
// This custom struct contains functions that must only be ran
|
||||
|
|
@ -175,6 +185,7 @@ impl BlockManager {
|
|||
let resync_queue = db
|
||||
.open_tree("block_local_resync_queue")
|
||||
.expect("Unable to open block_local_resync_queue tree");
|
||||
let resync_queue = SledCountedTree::new(resync_queue);
|
||||
|
||||
let endpoint = system
|
||||
.netapp
|
||||
|
|
@ -182,6 +193,8 @@ impl BlockManager {
|
|||
|
||||
let manager_locked = BlockManagerLocked();
|
||||
|
||||
let metrics = BlockManagerMetrics::new(resync_queue.clone());
|
||||
|
||||
let block_manager = Arc::new(Self {
|
||||
replication,
|
||||
data_dir,
|
||||
|
|
@ -192,6 +205,7 @@ impl BlockManager {
|
|||
system,
|
||||
endpoint,
|
||||
garage: ArcSwapOption::from(None),
|
||||
metrics,
|
||||
});
|
||||
block_manager.endpoint.set_handler(block_manager.clone());
|
||||
|
||||
|
|
@ -380,15 +394,36 @@ impl BlockManager {
|
|||
|
||||
/// Write a block to disk
|
||||
async fn write_block(&self, hash: &Hash, data: &DataBlock) -> Result<BlockRpc, Error> {
|
||||
self.mutation_lock
|
||||
let write_size = data.inner_buffer().len() as u64;
|
||||
|
||||
let res = self
|
||||
.mutation_lock
|
||||
.lock()
|
||||
.await
|
||||
.write_block(hash, data, self)
|
||||
.await
|
||||
.bound_record_duration(&self.metrics.block_write_duration)
|
||||
.await?;
|
||||
|
||||
self.metrics.bytes_written.add(write_size);
|
||||
|
||||
Ok(res)
|
||||
}
|
||||
|
||||
/// Read block from disk, verifying it's integrity
|
||||
async fn read_block(&self, hash: &Hash) -> Result<BlockRpc, Error> {
|
||||
let data = self
|
||||
.read_block_internal(hash)
|
||||
.bound_record_duration(&self.metrics.block_read_duration)
|
||||
.await?;
|
||||
|
||||
self.metrics
|
||||
.bytes_read
|
||||
.add(data.inner_buffer().len() as u64);
|
||||
|
||||
Ok(BlockRpc::PutBlock { hash: *hash, data })
|
||||
}
|
||||
|
||||
async fn read_block_internal(&self, hash: &Hash) -> Result<DataBlock, Error> {
|
||||
let mut path = self.block_path(hash);
|
||||
let compressed = match self.is_block_compressed(hash).await {
|
||||
Ok(c) => c,
|
||||
|
|
@ -414,6 +449,8 @@ impl BlockManager {
|
|||
};
|
||||
|
||||
if data.verify(*hash).is_err() {
|
||||
self.metrics.corruption_counter.add(1);
|
||||
|
||||
self.mutation_lock
|
||||
.lock()
|
||||
.await
|
||||
|
|
@ -423,7 +460,7 @@ impl BlockManager {
|
|||
return Err(Error::CorruptData(*hash));
|
||||
}
|
||||
|
||||
Ok(BlockRpc::PutBlock { hash: *hash, data })
|
||||
Ok(data)
|
||||
}
|
||||
|
||||
/// Check if this node should have a block, but don't actually have it
|
||||
|
|
@ -522,8 +559,30 @@ impl BlockManager {
|
|||
let now = now_msec();
|
||||
if now >= time_msec {
|
||||
let hash = Hash::try_from(&hash_bytes[..]).unwrap();
|
||||
let res = self.resync_block(&hash).await;
|
||||
|
||||
let tracer = opentelemetry::global::tracer("garage");
|
||||
let trace_id = gen_uuid();
|
||||
let span = tracer
|
||||
.span_builder("Resync block")
|
||||
.with_trace_id(
|
||||
opentelemetry::trace::TraceId::from_hex(&hex::encode(
|
||||
&trace_id.as_slice()[..16],
|
||||
))
|
||||
.unwrap(),
|
||||
)
|
||||
.with_attributes(vec![KeyValue::new("block", format!("{:?}", hash))])
|
||||
.start(&tracer);
|
||||
|
||||
let res = self
|
||||
.resync_block(&hash)
|
||||
.with_context(Context::current_with_span(span))
|
||||
.bound_record_duration(&self.metrics.resync_duration)
|
||||
.await;
|
||||
|
||||
self.metrics.resync_counter.add(1);
|
||||
|
||||
if let Err(e) = &res {
|
||||
self.metrics.resync_error_counter.add(1);
|
||||
warn!("Error when resyncing {:?}: {}", hash, e);
|
||||
self.put_to_resync(&hash, RESYNC_RETRY_DELAY)?;
|
||||
}
|
||||
|
|
@ -607,6 +666,12 @@ impl BlockManager {
|
|||
need_nodes.len()
|
||||
);
|
||||
|
||||
for node in need_nodes.iter() {
|
||||
self.metrics
|
||||
.resync_send_counter
|
||||
.add(1, &[KeyValue::new("to", format!("{:?}", node))]);
|
||||
}
|
||||
|
||||
let put_block_message = self.read_block(hash).await?;
|
||||
self.system
|
||||
.rpc
|
||||
|
|
@ -644,6 +709,9 @@ impl BlockManager {
|
|||
);
|
||||
|
||||
let block_data = self.rpc_get_raw_block(hash).await?;
|
||||
|
||||
self.metrics.resync_recv_counter.add(1);
|
||||
|
||||
self.write_block(hash, &block_data).await?;
|
||||
}
|
||||
|
||||
|
|
@ -819,6 +887,7 @@ impl BlockManagerLocked {
|
|||
path.set_extension("zst");
|
||||
}
|
||||
fs::remove_file(path).await?;
|
||||
mgr.metrics.delete_counter.add(1);
|
||||
}
|
||||
Ok(())
|
||||
}
|
||||
|
|
|
|||
|
|
@ -0,0 +1,95 @@
|
|||
use opentelemetry::{global, metrics::*};
|
||||
|
||||
use garage_util::sled_counter::SledCountedTree;
|
||||
|
||||
/// TableMetrics reference all counter used for metrics
|
||||
pub struct BlockManagerMetrics {
|
||||
pub(crate) _resync_queue_len: ValueObserver<u64>,
|
||||
|
||||
pub(crate) resync_counter: BoundCounter<u64>,
|
||||
pub(crate) resync_error_counter: BoundCounter<u64>,
|
||||
pub(crate) resync_duration: BoundValueRecorder<f64>,
|
||||
pub(crate) resync_send_counter: Counter<u64>,
|
||||
pub(crate) resync_recv_counter: BoundCounter<u64>,
|
||||
|
||||
pub(crate) bytes_read: BoundCounter<u64>,
|
||||
pub(crate) block_read_duration: BoundValueRecorder<f64>,
|
||||
pub(crate) bytes_written: BoundCounter<u64>,
|
||||
pub(crate) block_write_duration: BoundValueRecorder<f64>,
|
||||
pub(crate) delete_counter: BoundCounter<u64>,
|
||||
|
||||
pub(crate) corruption_counter: BoundCounter<u64>,
|
||||
}
|
||||
|
||||
impl BlockManagerMetrics {
|
||||
pub fn new(resync_queue: SledCountedTree) -> Self {
|
||||
let meter = global::meter("garage_model/block");
|
||||
Self {
|
||||
_resync_queue_len: meter
|
||||
.u64_value_observer("block.resync_queue_length", move |observer| {
|
||||
observer.observe(resync_queue.len() as u64, &[])
|
||||
})
|
||||
.with_description(
|
||||
"Number of block hashes queued for local check and possible resync",
|
||||
)
|
||||
.init(),
|
||||
|
||||
resync_counter: meter
|
||||
.u64_counter("block.resync_counter")
|
||||
.with_description("Number of calls to resync_block")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
resync_error_counter: meter
|
||||
.u64_counter("block.resync_error_counter")
|
||||
.with_description("Number of calls to resync_block that returned an error")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
resync_duration: meter
|
||||
.f64_value_recorder("block.resync_duration")
|
||||
.with_description("Duration of resync_block operations")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
resync_send_counter: meter
|
||||
.u64_counter("block.resync_send_counter")
|
||||
.with_description("Number of blocks sent to another node in resync operations")
|
||||
.init(),
|
||||
resync_recv_counter: meter
|
||||
.u64_counter("block.resync_recv_counter")
|
||||
.with_description("Number of blocks received from other nodes in resync operations")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
|
||||
bytes_read: meter
|
||||
.u64_counter("block.bytes_read")
|
||||
.with_description("Number of bytes read from disk")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
block_read_duration: meter
|
||||
.f64_value_recorder("block.read_duration")
|
||||
.with_description("Duration of block read operations")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
bytes_written: meter
|
||||
.u64_counter("block.bytes_written")
|
||||
.with_description("Number of bytes written to disk")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
block_write_duration: meter
|
||||
.f64_value_recorder("block.write_duration")
|
||||
.with_description("Duration of block write operations")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
delete_counter: meter
|
||||
.u64_counter("block.delete_counter")
|
||||
.with_description("Number of blocks deleted")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
|
||||
corruption_counter: meter
|
||||
.u64_counter("block.corruption_counter")
|
||||
.with_description("Data corruptions detected on block reads")
|
||||
.init()
|
||||
.bind(&[]),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
@ -1,5 +1,5 @@
|
|||
#[macro_use]
|
||||
extern crate log;
|
||||
extern crate tracing;
|
||||
|
||||
pub mod permission;
|
||||
|
||||
|
|
@ -11,6 +11,7 @@ pub mod object_table;
|
|||
pub mod version_table;
|
||||
|
||||
pub mod block;
|
||||
mod block_metrics;
|
||||
|
||||
pub mod garage;
|
||||
pub mod helper;
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
[package]
|
||||
name = "garage_rpc"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
authors = ["Alex Auvolat <alex@adnab.me>"]
|
||||
edition = "2018"
|
||||
license = "AGPL-3.0"
|
||||
|
|
@ -14,13 +14,14 @@ path = "lib.rs"
|
|||
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
|
||||
|
||||
[dependencies]
|
||||
garage_util = { version = "0.6.0", path = "../util" }
|
||||
garage_util = { version = "0.7.0", path = "../util" }
|
||||
garage_admin = { version = "0.7.0", path = "../admin" }
|
||||
|
||||
arc-swap = "1.0"
|
||||
bytes = "1.0"
|
||||
gethostname = "0.2"
|
||||
hex = "0.4"
|
||||
log = "0.4"
|
||||
tracing = "0.1.30"
|
||||
rand = "0.8"
|
||||
sodiumoxide = { version = "0.2.5-0", package = "kuska-sodiumoxide" }
|
||||
|
||||
|
|
@ -43,8 +44,11 @@ futures = "0.3"
|
|||
futures-util = "0.3"
|
||||
tokio = { version = "1.0", default-features = false, features = ["rt", "rt-multi-thread", "io-util", "net", "time", "macros", "sync", "signal", "fs"] }
|
||||
tokio-stream = { version = "0.1", features = ["net"] }
|
||||
opentelemetry = "0.17"
|
||||
|
||||
#netapp = { version = "0.3.0", git = "https://git.deuxfleurs.fr/lx/netapp" }
|
||||
netapp = "0.3.0"
|
||||
#netapp = { version = "0.4", path = "../../../netapp", features = ["telemetry"] }
|
||||
netapp = { version = "0.4", features = ["telemetry"] }
|
||||
|
||||
hyper = { version = "0.14", features = ["client", "http1", "runtime", "tcp"] }
|
||||
|
||||
|
|
|
|||
|
|
@ -139,10 +139,10 @@ pub async fn publish_consul_service(
|
|||
let resp = client.request(req).await?;
|
||||
debug!("Response of advertising to Consul: {:?}", resp);
|
||||
let resp_code = resp.status();
|
||||
let resp_bytes = &hyper::body::to_bytes(resp.into_body()).await?;
|
||||
debug!(
|
||||
"{}",
|
||||
std::str::from_utf8(&hyper::body::to_bytes(resp.into_body()).await?)
|
||||
.unwrap_or("<invalid utf8>")
|
||||
std::str::from_utf8(resp_bytes).unwrap_or("<invalid utf8>")
|
||||
);
|
||||
|
||||
if resp_code != StatusCode::OK {
|
||||
|
|
|
|||
|
|
@ -1,7 +1,7 @@
|
|||
//! Crate containing rpc related functions and types used in Garage
|
||||
|
||||
#[macro_use]
|
||||
extern crate log;
|
||||
extern crate tracing;
|
||||
|
||||
mod consul;
|
||||
mod kubernetes;
|
||||
|
|
@ -10,6 +10,7 @@ pub mod layout;
|
|||
pub mod ring;
|
||||
pub mod system;
|
||||
|
||||
mod metrics;
|
||||
pub mod rpc_helper;
|
||||
|
||||
pub use rpc_helper::*;
|
||||
|
|
|
|||
|
|
@ -0,0 +1,55 @@
|
|||
use std::sync::Arc;
|
||||
|
||||
use opentelemetry::{global, metrics::*};
|
||||
use tokio::sync::Semaphore;
|
||||
|
||||
/// TableMetrics reference all counter used for metrics
|
||||
pub struct RpcMetrics {
|
||||
pub(crate) _rpc_available_permits: ValueObserver<u64>,
|
||||
|
||||
pub(crate) rpc_counter: Counter<u64>,
|
||||
pub(crate) rpc_timeout_counter: Counter<u64>,
|
||||
pub(crate) rpc_netapp_error_counter: Counter<u64>,
|
||||
pub(crate) rpc_garage_error_counter: Counter<u64>,
|
||||
|
||||
pub(crate) rpc_duration: ValueRecorder<f64>,
|
||||
pub(crate) rpc_queueing_time: ValueRecorder<f64>,
|
||||
}
|
||||
impl RpcMetrics {
|
||||
pub fn new(sem: Arc<Semaphore>) -> Self {
|
||||
let meter = global::meter("garage_rpc");
|
||||
RpcMetrics {
|
||||
_rpc_available_permits: meter
|
||||
.u64_value_observer("rpc.available_permits", move |observer| {
|
||||
observer.observe(sem.available_permits() as u64, &[])
|
||||
})
|
||||
.with_description("Number of available RPC permits")
|
||||
.init(),
|
||||
|
||||
rpc_counter: meter
|
||||
.u64_counter("rpc.request_counter")
|
||||
.with_description("Number of RPC requests emitted")
|
||||
.init(),
|
||||
rpc_timeout_counter: meter
|
||||
.u64_counter("rpc.timeout_counter")
|
||||
.with_description("Number of RPC timeouts")
|
||||
.init(),
|
||||
rpc_netapp_error_counter: meter
|
||||
.u64_counter("rpc.netapp_error_counter")
|
||||
.with_description("Number of communication errors (errors in the Netapp library)")
|
||||
.init(),
|
||||
rpc_garage_error_counter: meter
|
||||
.u64_counter("rpc.garage_error_counter")
|
||||
.with_description("Number of RPC errors (errors happening when handling the RPC)")
|
||||
.init(),
|
||||
rpc_duration: meter
|
||||
.f64_value_recorder("rpc.duration")
|
||||
.with_description("Duration of RPCs")
|
||||
.init(),
|
||||
rpc_queueing_time: meter
|
||||
.f64_value_recorder("rpc.queueing_time")
|
||||
.with_description("Time RPC requests were queued for before being sent")
|
||||
.init(),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
@ -9,6 +9,12 @@ use futures_util::future::FutureExt;
|
|||
use tokio::select;
|
||||
use tokio::sync::{watch, Semaphore};
|
||||
|
||||
use opentelemetry::KeyValue;
|
||||
use opentelemetry::{
|
||||
trace::{FutureExt as OtelFutureExt, Span, TraceContextExt, Tracer},
|
||||
Context,
|
||||
};
|
||||
|
||||
pub use netapp::endpoint::{Endpoint, EndpointHandler, Message as Rpc};
|
||||
use netapp::peering::fullmesh::FullMeshPeeringStrategy;
|
||||
pub use netapp::proto::*;
|
||||
|
|
@ -17,7 +23,9 @@ pub use netapp::{NetApp, NodeID};
|
|||
use garage_util::background::BackgroundRunner;
|
||||
use garage_util::data::*;
|
||||
use garage_util::error::Error;
|
||||
use garage_util::metrics::RecordDuration;
|
||||
|
||||
use crate::metrics::RpcMetrics;
|
||||
use crate::ring::Ring;
|
||||
|
||||
const DEFAULT_TIMEOUT: Duration = Duration::from_secs(10);
|
||||
|
|
@ -76,7 +84,8 @@ struct RpcHelperInner {
|
|||
fullmesh: Arc<FullMeshPeeringStrategy>,
|
||||
background: Arc<BackgroundRunner>,
|
||||
ring: watch::Receiver<Arc<Ring>>,
|
||||
request_buffer_semaphore: Semaphore,
|
||||
request_buffer_semaphore: Arc<Semaphore>,
|
||||
metrics: RpcMetrics,
|
||||
}
|
||||
|
||||
impl RpcHelper {
|
||||
|
|
@ -86,12 +95,17 @@ impl RpcHelper {
|
|||
background: Arc<BackgroundRunner>,
|
||||
ring: watch::Receiver<Arc<Ring>>,
|
||||
) -> Self {
|
||||
let sem = Arc::new(Semaphore::new(REQUEST_BUFFER_SIZE));
|
||||
|
||||
let metrics = RpcMetrics::new(sem.clone());
|
||||
|
||||
Self(Arc::new(RpcHelperInner {
|
||||
our_node_id,
|
||||
fullmesh,
|
||||
background,
|
||||
ring,
|
||||
request_buffer_semaphore: Semaphore::new(REQUEST_BUFFER_SIZE),
|
||||
request_buffer_semaphore: sem,
|
||||
metrics,
|
||||
}))
|
||||
}
|
||||
|
||||
|
|
@ -120,21 +134,45 @@ impl RpcHelper {
|
|||
M: Rpc<Response = Result<S, Error>>,
|
||||
H: EndpointHandler<M>,
|
||||
{
|
||||
let metric_tags = [
|
||||
KeyValue::new("rpc_endpoint", endpoint.path().to_string()),
|
||||
KeyValue::new("from", format!("{:?}", self.0.our_node_id)),
|
||||
KeyValue::new("to", format!("{:?}", to)),
|
||||
];
|
||||
|
||||
let msg_size = rmp_to_vec_all_named(&msg)?.len() as u32;
|
||||
let permit = self
|
||||
.0
|
||||
.request_buffer_semaphore
|
||||
.acquire_many(msg_size)
|
||||
.record_duration(&self.0.metrics.rpc_queueing_time, &metric_tags)
|
||||
.await?;
|
||||
|
||||
self.0.metrics.rpc_counter.add(1, &metric_tags);
|
||||
|
||||
let node_id = to.into();
|
||||
let rpc_call = endpoint
|
||||
.call(&node_id, msg, strat.rs_priority)
|
||||
.record_duration(&self.0.metrics.rpc_duration, &metric_tags);
|
||||
|
||||
select! {
|
||||
res = endpoint.call(&node_id, &msg, strat.rs_priority) => {
|
||||
res = rpc_call => {
|
||||
drop(permit);
|
||||
Ok(res??)
|
||||
|
||||
if res.is_err() {
|
||||
self.0.metrics.rpc_netapp_error_counter.add(1, &metric_tags);
|
||||
}
|
||||
let res = res?;
|
||||
|
||||
if res.is_err() {
|
||||
self.0.metrics.rpc_garage_error_counter.add(1, &metric_tags);
|
||||
}
|
||||
|
||||
Ok(res?)
|
||||
}
|
||||
_ = tokio::time::sleep(strat.rs_timeout) => {
|
||||
drop(permit);
|
||||
self.0.metrics.rpc_timeout_counter.add(1, &metric_tags);
|
||||
Err(Error::Timeout)
|
||||
}
|
||||
}
|
||||
|
|
@ -195,7 +233,47 @@ impl RpcHelper {
|
|||
where
|
||||
M: Rpc<Response = Result<S, Error>> + 'static,
|
||||
H: EndpointHandler<M> + 'static,
|
||||
S: Send,
|
||||
S: Send + 'static,
|
||||
{
|
||||
let quorum = strategy.rs_quorum.unwrap_or(to.len());
|
||||
|
||||
let tracer = opentelemetry::global::tracer("garage");
|
||||
let span_name = if strategy.rs_interrupt_after_quorum {
|
||||
format!("RPC {} to {} of {}", endpoint.path(), quorum, to.len())
|
||||
} else {
|
||||
format!(
|
||||
"RPC {} to {} (quorum {})",
|
||||
endpoint.path(),
|
||||
to.len(),
|
||||
quorum
|
||||
)
|
||||
};
|
||||
let mut span = tracer.start(span_name);
|
||||
span.set_attribute(KeyValue::new("from", format!("{:?}", self.0.our_node_id)));
|
||||
span.set_attribute(KeyValue::new("to", format!("{:?}", to)));
|
||||
span.set_attribute(KeyValue::new("quorum", quorum as i64));
|
||||
span.set_attribute(KeyValue::new(
|
||||
"interrupt_after_quorum",
|
||||
strategy.rs_interrupt_after_quorum.to_string(),
|
||||
));
|
||||
|
||||
self.try_call_many_internal(endpoint, to, msg, strategy, quorum)
|
||||
.with_context(Context::current_with_span(span))
|
||||
.await
|
||||
}
|
||||
|
||||
async fn try_call_many_internal<M, H, S>(
|
||||
&self,
|
||||
endpoint: &Arc<Endpoint<M, H>>,
|
||||
to: &[Uuid],
|
||||
msg: M,
|
||||
strategy: RequestStrategy,
|
||||
quorum: usize,
|
||||
) -> Result<Vec<S>, Error>
|
||||
where
|
||||
M: Rpc<Response = Result<S, Error>> + 'static,
|
||||
H: EndpointHandler<M> + 'static,
|
||||
S: Send + 'static,
|
||||
{
|
||||
let msg = Arc::new(msg);
|
||||
|
||||
|
|
@ -210,7 +288,6 @@ impl RpcHelper {
|
|||
self2.call_arc(&endpoint2, to, msg, strategy).await
|
||||
})
|
||||
});
|
||||
let quorum = strategy.rs_quorum.unwrap_or(to.len());
|
||||
|
||||
// Vectors in which success results and errors will be collected
|
||||
let mut successes = vec![];
|
||||
|
|
@ -274,8 +351,12 @@ impl RpcHelper {
|
|||
// If the current set of requests that are running is not enough to possibly
|
||||
// reach quorum, start some new requests.
|
||||
while successes.len() + resp_stream.len() < quorum {
|
||||
if let Some((_, _, _, _to, fut)) = requests.next() {
|
||||
resp_stream.push(fut);
|
||||
if let Some((_, _, _, req_to, fut)) = requests.next() {
|
||||
let tracer = opentelemetry::global::tracer("garage");
|
||||
let span = tracer.start(format!("RPC to {:?}", req_to));
|
||||
resp_stream.push(tokio::spawn(
|
||||
fut.with_context(Context::current_with_span(span)),
|
||||
));
|
||||
} else {
|
||||
// If we have no request to add, we know that we won't ever
|
||||
// reach quorum: bail out now.
|
||||
|
|
@ -285,7 +366,7 @@ impl RpcHelper {
|
|||
assert!(!resp_stream.is_empty()); // because of loop invariants
|
||||
|
||||
// Wait for one request to terminate
|
||||
match resp_stream.next().await.unwrap() {
|
||||
match resp_stream.next().await.unwrap().unwrap() {
|
||||
Ok(msg) => {
|
||||
successes.push(msg);
|
||||
}
|
||||
|
|
|
|||
|
|
@ -38,6 +38,9 @@ const DISCOVERY_INTERVAL: Duration = Duration::from_secs(60);
|
|||
const STATUS_EXCHANGE_INTERVAL: Duration = Duration::from_secs(10);
|
||||
const PING_TIMEOUT: Duration = Duration::from_secs(2);
|
||||
|
||||
/// Version tag used for version check upon Netapp connection
|
||||
pub const GARAGE_VERSION_TAG: u64 = 0x6761726167650007; // garage 0x0007
|
||||
|
||||
/// RPC endpoint used for calls related to membership
|
||||
pub const SYSTEM_RPC_PATH: &str = "garage_rpc/membership.rs/SystemRpc";
|
||||
|
||||
|
|
@ -194,7 +197,10 @@ impl System {
|
|||
) -> Arc<Self> {
|
||||
let node_key =
|
||||
gen_node_key(&config.metadata_dir).expect("Unable to read or generate node ID");
|
||||
info!("Node public key: {}", hex::encode(&node_key.public_key()));
|
||||
info!(
|
||||
"Node ID of this node: {}",
|
||||
hex::encode(&node_key.public_key()[..8])
|
||||
);
|
||||
|
||||
let persist_cluster_layout = Persister::new(&config.metadata_dir, "cluster_layout");
|
||||
let persist_peer_list = Persister::new(&config.metadata_dir, "peer_list");
|
||||
|
|
@ -222,13 +228,7 @@ impl System {
|
|||
let ring = Ring::new(cluster_layout, replication_factor);
|
||||
let (update_ring, ring) = watch::channel(Arc::new(ring));
|
||||
|
||||
if let Some(addr) = config.rpc_public_addr {
|
||||
println!("{}@{}", hex::encode(&node_key.public_key()), addr);
|
||||
} else {
|
||||
println!("{}", hex::encode(&node_key.public_key()));
|
||||
}
|
||||
|
||||
let netapp = NetApp::new(network_key, node_key);
|
||||
let netapp = NetApp::new(GARAGE_VERSION_TAG, network_key, node_key);
|
||||
let fullmesh = FullMeshPeeringStrategy::new(
|
||||
netapp.clone(),
|
||||
config.bootstrap_peers.clone(),
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
[package]
|
||||
name = "garage_table"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
authors = ["Alex Auvolat <alex@adnab.me>"]
|
||||
edition = "2018"
|
||||
license = "AGPL-3.0"
|
||||
|
|
@ -14,13 +14,15 @@ path = "lib.rs"
|
|||
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
|
||||
|
||||
[dependencies]
|
||||
garage_rpc = { version = "0.6.0", path = "../rpc" }
|
||||
garage_util = { version = "0.6.0", path = "../util" }
|
||||
garage_rpc = { version = "0.7.0", path = "../rpc" }
|
||||
garage_util = { version = "0.7.0", path = "../util" }
|
||||
|
||||
opentelemetry = "0.17"
|
||||
|
||||
async-trait = "0.1.7"
|
||||
bytes = "1.0"
|
||||
hexdump = "0.1"
|
||||
log = "0.4"
|
||||
tracing = "0.1.30"
|
||||
rand = "0.8"
|
||||
|
||||
sled = "0.34"
|
||||
|
|
|
|||
|
|
@ -1,18 +1,19 @@
|
|||
use core::borrow::Borrow;
|
||||
use std::sync::Arc;
|
||||
|
||||
use log::warn;
|
||||
use serde_bytes::ByteBuf;
|
||||
use sled::Transactional;
|
||||
use tokio::sync::Notify;
|
||||
|
||||
use garage_util::data::*;
|
||||
use garage_util::error::*;
|
||||
use garage_util::sled_counter::SledCountedTree;
|
||||
|
||||
use garage_rpc::system::System;
|
||||
|
||||
use crate::crdt::Crdt;
|
||||
use crate::gc::GcTodoEntry;
|
||||
use crate::metrics::*;
|
||||
use crate::replication::*;
|
||||
use crate::schema::*;
|
||||
|
||||
|
|
@ -27,7 +28,9 @@ pub struct TableData<F: TableSchema, R: TableReplication> {
|
|||
pub(crate) merkle_tree: sled::Tree,
|
||||
pub(crate) merkle_todo: sled::Tree,
|
||||
pub(crate) merkle_todo_notify: Notify,
|
||||
pub(crate) gc_todo: sled::Tree,
|
||||
pub(crate) gc_todo: SledCountedTree,
|
||||
|
||||
pub(crate) metrics: TableMetrics,
|
||||
}
|
||||
|
||||
impl<F, R> TableData<F, R>
|
||||
|
|
@ -50,6 +53,9 @@ where
|
|||
let gc_todo = db
|
||||
.open_tree(&format!("{}:gc_todo_v2", F::TABLE_NAME))
|
||||
.expect("Unable to open DB tree");
|
||||
let gc_todo = SledCountedTree::new(gc_todo);
|
||||
|
||||
let metrics = TableMetrics::new(F::TABLE_NAME, merkle_todo.clone(), gc_todo.clone());
|
||||
|
||||
Arc::new(Self {
|
||||
system,
|
||||
|
|
@ -60,6 +66,7 @@ where
|
|||
merkle_todo,
|
||||
merkle_todo_notify: Notify::new(),
|
||||
gc_todo,
|
||||
metrics,
|
||||
})
|
||||
}
|
||||
|
||||
|
|
@ -165,6 +172,8 @@ where
|
|||
})?;
|
||||
|
||||
if let Some((old_entry, new_entry, new_bytes_hash)) = changed {
|
||||
self.metrics.internal_update_counter.add(1);
|
||||
|
||||
let is_tombstone = new_entry.is_tombstone();
|
||||
self.instance.updated(old_entry, Some(new_entry));
|
||||
self.merkle_todo_notify.notify_one();
|
||||
|
|
@ -199,6 +208,8 @@ where
|
|||
})?;
|
||||
|
||||
if removed {
|
||||
self.metrics.internal_delete_counter.add(1);
|
||||
|
||||
let old_entry = self.decode_entry(v)?;
|
||||
self.instance.updated(Some(old_entry), None);
|
||||
self.merkle_todo_notify.notify_one();
|
||||
|
|
|
|||
|
|
@ -14,6 +14,7 @@ use tokio::sync::watch;
|
|||
|
||||
use garage_util::data::*;
|
||||
use garage_util::error::*;
|
||||
use garage_util::sled_counter::SledCountedTree;
|
||||
use garage_util::time::*;
|
||||
|
||||
use garage_rpc::system::System;
|
||||
|
|
@ -362,7 +363,7 @@ impl GcTodoEntry {
|
|||
}
|
||||
|
||||
/// Saves the GcTodoEntry in the gc_todo tree
|
||||
pub(crate) fn save(&self, gc_todo_tree: &sled::Tree) -> Result<(), Error> {
|
||||
pub(crate) fn save(&self, gc_todo_tree: &SledCountedTree) -> Result<(), Error> {
|
||||
gc_todo_tree.insert(self.todo_table_key(), self.value_hash.as_slice())?;
|
||||
Ok(())
|
||||
}
|
||||
|
|
@ -372,7 +373,7 @@ impl GcTodoEntry {
|
|||
/// This is usefull to remove a todo entry only under the condition
|
||||
/// that it has not changed since the time it was read, i.e.
|
||||
/// what we have to do is still the same
|
||||
pub(crate) fn remove_if_equal(&self, gc_todo_tree: &sled::Tree) -> Result<(), Error> {
|
||||
pub(crate) fn remove_if_equal(&self, gc_todo_tree: &SledCountedTree) -> Result<(), Error> {
|
||||
let _ = gc_todo_tree.compare_and_swap::<_, _, Vec<u8>>(
|
||||
&self.todo_table_key()[..],
|
||||
Some(self.value_hash),
|
||||
|
|
|
|||
|
|
@ -2,8 +2,9 @@
|
|||
#![allow(clippy::comparison_chain)]
|
||||
|
||||
#[macro_use]
|
||||
extern crate log;
|
||||
extern crate tracing;
|
||||
|
||||
mod metrics;
|
||||
pub mod schema;
|
||||
pub mod util;
|
||||
|
||||
|
|
|
|||
|
|
@ -3,7 +3,6 @@ use std::time::Duration;
|
|||
|
||||
use futures::select;
|
||||
use futures_util::future::*;
|
||||
use log::{debug, warn};
|
||||
use serde::{Deserialize, Serialize};
|
||||
use sled::transaction::{
|
||||
ConflictableTransactionError, ConflictableTransactionResult, TransactionalTree,
|
||||
|
|
|
|||
|
|
@ -0,0 +1,96 @@
|
|||
use opentelemetry::{global, metrics::*, KeyValue};
|
||||
|
||||
use garage_util::sled_counter::SledCountedTree;
|
||||
|
||||
/// TableMetrics reference all counter used for metrics
|
||||
pub struct TableMetrics {
|
||||
pub(crate) _merkle_todo_len: ValueObserver<u64>,
|
||||
pub(crate) _gc_todo_len: ValueObserver<u64>,
|
||||
|
||||
pub(crate) get_request_counter: BoundCounter<u64>,
|
||||
pub(crate) get_request_duration: BoundValueRecorder<f64>,
|
||||
pub(crate) put_request_counter: BoundCounter<u64>,
|
||||
pub(crate) put_request_duration: BoundValueRecorder<f64>,
|
||||
|
||||
pub(crate) internal_update_counter: BoundCounter<u64>,
|
||||
pub(crate) internal_delete_counter: BoundCounter<u64>,
|
||||
|
||||
pub(crate) sync_items_sent: Counter<u64>,
|
||||
pub(crate) sync_items_received: Counter<u64>,
|
||||
}
|
||||
impl TableMetrics {
|
||||
pub fn new(
|
||||
table_name: &'static str,
|
||||
merkle_todo: sled::Tree,
|
||||
gc_todo: SledCountedTree,
|
||||
) -> Self {
|
||||
let meter = global::meter(table_name);
|
||||
TableMetrics {
|
||||
_merkle_todo_len: meter
|
||||
.u64_value_observer(
|
||||
"table.merkle_updater_todo_queue_length",
|
||||
move |observer| {
|
||||
observer.observe(
|
||||
merkle_todo.len() as u64,
|
||||
&[KeyValue::new("table_name", table_name)],
|
||||
)
|
||||
},
|
||||
)
|
||||
.with_description("Merkle tree updater TODO queue length")
|
||||
.init(),
|
||||
_gc_todo_len: meter
|
||||
.u64_value_observer(
|
||||
"table.gc_todo_queue_length",
|
||||
move |observer| {
|
||||
observer.observe(
|
||||
gc_todo.len() as u64,
|
||||
&[KeyValue::new("table_name", table_name)],
|
||||
)
|
||||
},
|
||||
)
|
||||
.with_description("Table garbage collector TODO queue length")
|
||||
.init(),
|
||||
|
||||
get_request_counter: meter
|
||||
.u64_counter("table.get_request_counter")
|
||||
.with_description("Number of get/get_range requests internally made on this table")
|
||||
.init()
|
||||
.bind(&[KeyValue::new("table_name", table_name)]),
|
||||
get_request_duration: meter
|
||||
.f64_value_recorder("table.get_request_duration")
|
||||
.with_description("Duration of get/get_range requests internally made on this table, in seconds")
|
||||
.init()
|
||||
.bind(&[KeyValue::new("table_name", table_name)]),
|
||||
put_request_counter: meter
|
||||
.u64_counter("table.put_request_counter")
|
||||
.with_description("Number of insert/insert_many requests internally made on this table")
|
||||
.init()
|
||||
.bind(&[KeyValue::new("table_name", table_name)]),
|
||||
put_request_duration: meter
|
||||
.f64_value_recorder("table.put_request_duration")
|
||||
.with_description("Duration of insert/insert_many requests internally made on this table, in seconds")
|
||||
.init()
|
||||
.bind(&[KeyValue::new("table_name", table_name)]),
|
||||
|
||||
internal_update_counter: meter
|
||||
.u64_counter("table.internal_update_counter")
|
||||
.with_description("Number of value updates where the value actually changes (includes creation of new key and update of existing key)")
|
||||
.init()
|
||||
.bind(&[KeyValue::new("table_name", table_name)]),
|
||||
internal_delete_counter: meter
|
||||
.u64_counter("table.internal_delete_counter")
|
||||
.with_description("Number of value deletions in the tree (due to GC or repartitioning)")
|
||||
.init()
|
||||
.bind(&[KeyValue::new("table_name", table_name)]),
|
||||
|
||||
sync_items_sent: meter
|
||||
.u64_counter("table.sync_items_sent")
|
||||
.with_description("Number of data items sent to other nodes during resync procedures")
|
||||
.init(),
|
||||
sync_items_received: meter
|
||||
.u64_counter("table.sync_items_received")
|
||||
.with_description("Number of data items received from other nodes during resync procedures")
|
||||
.init(),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
@ -6,6 +6,7 @@ use async_trait::async_trait;
|
|||
use futures::select;
|
||||
use futures_util::future::*;
|
||||
use futures_util::stream::*;
|
||||
use opentelemetry::KeyValue;
|
||||
use rand::Rng;
|
||||
use serde::{Deserialize, Serialize};
|
||||
use serde_bytes::ByteBuf;
|
||||
|
|
@ -312,6 +313,16 @@ where
|
|||
) -> Result<(), Error> {
|
||||
let values = items.iter().map(|(_k, v)| v.clone()).collect::<Vec<_>>();
|
||||
|
||||
for to in nodes.iter() {
|
||||
self.data.metrics.sync_items_sent.add(
|
||||
values.len() as u64,
|
||||
&[
|
||||
KeyValue::new("table_name", F::TABLE_NAME),
|
||||
KeyValue::new("to", format!("{:?}", to)),
|
||||
],
|
||||
);
|
||||
}
|
||||
|
||||
self.system
|
||||
.rpc
|
||||
.try_call_many(
|
||||
|
|
@ -500,6 +511,14 @@ where
|
|||
.map(|x| Arc::new(ByteBuf::from(x)))
|
||||
.collect::<Vec<_>>();
|
||||
|
||||
self.data.metrics.sync_items_sent.add(
|
||||
values.len() as u64,
|
||||
&[
|
||||
KeyValue::new("table_name", F::TABLE_NAME),
|
||||
KeyValue::new("to", format!("{:?}", who)),
|
||||
],
|
||||
);
|
||||
|
||||
let rpc_resp = self
|
||||
.system
|
||||
.rpc
|
||||
|
|
@ -527,7 +546,7 @@ where
|
|||
F: TableSchema + 'static,
|
||||
R: TableReplication + 'static,
|
||||
{
|
||||
async fn handle(self: &Arc<Self>, message: &SyncRpc, _from: NodeID) -> Result<SyncRpc, Error> {
|
||||
async fn handle(self: &Arc<Self>, message: &SyncRpc, from: NodeID) -> Result<SyncRpc, Error> {
|
||||
match message {
|
||||
SyncRpc::RootCkHash(range, h) => {
|
||||
let (_root_ck_key, root_ck) = self.get_root_ck(*range)?;
|
||||
|
|
@ -539,6 +558,17 @@ where
|
|||
Ok(SyncRpc::Node(k.clone(), node))
|
||||
}
|
||||
SyncRpc::Items(items) => {
|
||||
self.data.metrics.sync_items_received.add(
|
||||
items.len() as u64,
|
||||
&[
|
||||
KeyValue::new("table_name", F::TABLE_NAME),
|
||||
KeyValue::new(
|
||||
"from",
|
||||
format!("{:?}", Uuid::try_from(from.as_ref()).unwrap()),
|
||||
),
|
||||
],
|
||||
);
|
||||
|
||||
self.data.update_many(items)?;
|
||||
Ok(SyncRpc::Ok)
|
||||
}
|
||||
|
|
|
|||
|
|
@ -7,8 +7,14 @@ use futures::stream::*;
|
|||
use serde::{Deserialize, Serialize};
|
||||
use serde_bytes::ByteBuf;
|
||||
|
||||
use opentelemetry::{
|
||||
trace::{FutureExt, TraceContextExt, Tracer},
|
||||
Context,
|
||||
};
|
||||
|
||||
use garage_util::data::*;
|
||||
use garage_util::error::Error;
|
||||
use garage_util::metrics::RecordDuration;
|
||||
|
||||
use garage_rpc::system::System;
|
||||
use garage_rpc::*;
|
||||
|
|
@ -81,6 +87,20 @@ where
|
|||
}
|
||||
|
||||
pub async fn insert(&self, e: &F::E) -> Result<(), Error> {
|
||||
let tracer = opentelemetry::global::tracer("garage_table");
|
||||
let span = tracer.start(format!("{} insert", F::TABLE_NAME));
|
||||
|
||||
self.insert_internal(e)
|
||||
.bound_record_duration(&self.data.metrics.put_request_duration)
|
||||
.with_context(Context::current_with_span(span))
|
||||
.await?;
|
||||
|
||||
self.data.metrics.put_request_counter.add(1);
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
async fn insert_internal(&self, e: &F::E) -> Result<(), Error> {
|
||||
let hash = e.partition_key().hash();
|
||||
let who = self.data.replication.write_nodes(&hash);
|
||||
//eprintln!("insert who: {:?}", who);
|
||||
|
|
@ -99,10 +119,25 @@ where
|
|||
.with_timeout(TABLE_RPC_TIMEOUT),
|
||||
)
|
||||
.await?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
pub async fn insert_many(&self, entries: &[F::E]) -> Result<(), Error> {
|
||||
let tracer = opentelemetry::global::tracer("garage_table");
|
||||
let span = tracer.start(format!("{} insert_many {}", F::TABLE_NAME, entries.len()));
|
||||
|
||||
self.insert_many_internal(entries)
|
||||
.bound_record_duration(&self.data.metrics.put_request_duration)
|
||||
.with_context(Context::current_with_span(span))
|
||||
.await?;
|
||||
|
||||
self.data.metrics.put_request_counter.add(1);
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
async fn insert_many_internal(&self, entries: &[F::E]) -> Result<(), Error> {
|
||||
let mut call_list: HashMap<_, Vec<_>> = HashMap::new();
|
||||
|
||||
for entry in entries.iter() {
|
||||
|
|
@ -148,10 +183,28 @@ where
|
|||
self: &Arc<Self>,
|
||||
partition_key: &F::P,
|
||||
sort_key: &F::S,
|
||||
) -> Result<Option<F::E>, Error> {
|
||||
let tracer = opentelemetry::global::tracer("garage_table");
|
||||
let span = tracer.start(format!("{} get", F::TABLE_NAME));
|
||||
|
||||
let res = self
|
||||
.get_internal(partition_key, sort_key)
|
||||
.bound_record_duration(&self.data.metrics.get_request_duration)
|
||||
.with_context(Context::current_with_span(span))
|
||||
.await?;
|
||||
|
||||
self.data.metrics.get_request_counter.add(1);
|
||||
|
||||
Ok(res)
|
||||
}
|
||||
|
||||
async fn get_internal(
|
||||
self: &Arc<Self>,
|
||||
partition_key: &F::P,
|
||||
sort_key: &F::S,
|
||||
) -> Result<Option<F::E>, Error> {
|
||||
let hash = partition_key.hash();
|
||||
let who = self.data.replication.read_nodes(&hash);
|
||||
//eprintln!("get who: {:?}", who);
|
||||
|
||||
let rpc = TableRpc::<F>::ReadEntry(partition_key.clone(), sort_key.clone());
|
||||
let resps = self
|
||||
|
|
@ -198,6 +251,7 @@ where
|
|||
.spawn_cancellable(async move { self2.repair_on_read(&who[..], ent2).await });
|
||||
}
|
||||
}
|
||||
|
||||
Ok(ret)
|
||||
}
|
||||
|
||||
|
|
@ -207,6 +261,27 @@ where
|
|||
begin_sort_key: Option<F::S>,
|
||||
filter: Option<F::Filter>,
|
||||
limit: usize,
|
||||
) -> Result<Vec<F::E>, Error> {
|
||||
let tracer = opentelemetry::global::tracer("garage_table");
|
||||
let span = tracer.start(format!("{} get_range", F::TABLE_NAME));
|
||||
|
||||
let res = self
|
||||
.get_range_internal(partition_key, begin_sort_key, filter, limit)
|
||||
.bound_record_duration(&self.data.metrics.get_request_duration)
|
||||
.with_context(Context::current_with_span(span))
|
||||
.await?;
|
||||
|
||||
self.data.metrics.get_request_counter.add(1);
|
||||
|
||||
Ok(res)
|
||||
}
|
||||
|
||||
async fn get_range_internal(
|
||||
self: &Arc<Self>,
|
||||
partition_key: &F::P,
|
||||
begin_sort_key: Option<F::S>,
|
||||
filter: Option<F::Filter>,
|
||||
limit: usize,
|
||||
) -> Result<Vec<F::E>, Error> {
|
||||
let hash = partition_key.hash();
|
||||
let who = self.data.replication.read_nodes(&hash);
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
[package]
|
||||
name = "garage_util"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
authors = ["Alex Auvolat <alex@adnab.me>"]
|
||||
edition = "2018"
|
||||
license = "AGPL-3.0"
|
||||
|
|
@ -18,7 +18,7 @@ blake2 = "0.9"
|
|||
err-derive = "0.3"
|
||||
xxhash-rust = { version = "0.8", default-features = false, features = ["xxh3"] }
|
||||
hex = "0.4"
|
||||
log = "0.4"
|
||||
tracing = "0.1.30"
|
||||
rand = "0.8"
|
||||
sha2 = "0.9"
|
||||
|
||||
|
|
@ -34,10 +34,13 @@ futures = "0.3"
|
|||
tokio = { version = "1.0", default-features = false, features = ["rt", "rt-multi-thread", "io-util", "net", "time", "macros", "sync", "signal", "fs"] }
|
||||
|
||||
#netapp = { version = "0.3.0", git = "https://git.deuxfleurs.fr/lx/netapp" }
|
||||
netapp = "0.3.0"
|
||||
#netapp = { version = "0.4", path = "../../../netapp" }
|
||||
netapp = "0.4"
|
||||
|
||||
http = "0.2"
|
||||
hyper = "0.14"
|
||||
|
||||
kube = { version = "0.62", features = ["runtime", "derive"] }
|
||||
k8s-openapi = { version = "0.13", features = ["v1_22"] }
|
||||
|
||||
opentelemetry = "0.17"
|
||||
|
|
|
|||
|
|
@ -73,6 +73,10 @@ pub struct Config {
|
|||
|
||||
/// Configuration for serving files as normal web server
|
||||
pub s3_web: WebConfig,
|
||||
|
||||
/// Configuration for the admin API endpoint
|
||||
#[serde(default = "Default::default")]
|
||||
pub admin: AdminConfig,
|
||||
}
|
||||
|
||||
/// Configuration for S3 api
|
||||
|
|
@ -96,6 +100,15 @@ pub struct WebConfig {
|
|||
pub root_domain: String,
|
||||
}
|
||||
|
||||
/// Configuration for the admin and monitoring HTTP API
|
||||
#[derive(Deserialize, Debug, Clone, Default)]
|
||||
pub struct AdminConfig {
|
||||
/// Address and port to bind for admin API serving
|
||||
pub api_bind_addr: Option<SocketAddr>,
|
||||
/// OTLP server to where to export traces
|
||||
pub trace_sink: Option<String>,
|
||||
}
|
||||
|
||||
fn default_sled_cache_capacity() -> u64 {
|
||||
128 * 1024 * 1024
|
||||
}
|
||||
|
|
|
|||
|
|
@ -22,7 +22,7 @@ impl std::convert::AsRef<[u8]> for FixedBytes32 {
|
|||
|
||||
impl fmt::Debug for FixedBytes32 {
|
||||
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
|
||||
write!(f, "{}…", hex::encode(&self.0[..8]))
|
||||
write!(f, "{}", hex::encode(&self.0[..8]))
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -1,14 +1,16 @@
|
|||
//! Crate containing common functions and types used in Garage
|
||||
|
||||
#[macro_use]
|
||||
extern crate log;
|
||||
extern crate tracing;
|
||||
|
||||
pub mod background;
|
||||
pub mod config;
|
||||
pub mod crdt;
|
||||
pub mod data;
|
||||
pub mod error;
|
||||
pub mod metrics;
|
||||
pub mod persister;
|
||||
pub mod sled_counter;
|
||||
pub mod time;
|
||||
pub mod token_bucket;
|
||||
pub mod tranquilizer;
|
||||
|
|
|
|||
|
|
@ -0,0 +1,57 @@
|
|||
use std::time::SystemTime;
|
||||
|
||||
use futures::{future::BoxFuture, Future, FutureExt};
|
||||
use rand::Rng;
|
||||
|
||||
use opentelemetry::{metrics::*, trace::TraceId, KeyValue};
|
||||
|
||||
pub trait RecordDuration<'a>: 'a {
|
||||
type Output;
|
||||
|
||||
fn record_duration(
|
||||
self,
|
||||
r: &'a ValueRecorder<f64>,
|
||||
attributes: &'a [KeyValue],
|
||||
) -> BoxFuture<'a, Self::Output>;
|
||||
fn bound_record_duration(self, r: &'a BoundValueRecorder<f64>) -> BoxFuture<'a, Self::Output>;
|
||||
}
|
||||
|
||||
impl<'a, T, O> RecordDuration<'a> for T
|
||||
where
|
||||
T: Future<Output = O> + Send + 'a,
|
||||
{
|
||||
type Output = O;
|
||||
|
||||
fn record_duration(
|
||||
self,
|
||||
r: &'a ValueRecorder<f64>,
|
||||
attributes: &'a [KeyValue],
|
||||
) -> BoxFuture<'a, Self::Output> {
|
||||
async move {
|
||||
let request_start = SystemTime::now();
|
||||
let res = self.await;
|
||||
r.record(
|
||||
request_start.elapsed().map_or(0.0, |d| d.as_secs_f64()),
|
||||
attributes,
|
||||
);
|
||||
res
|
||||
}
|
||||
.boxed()
|
||||
}
|
||||
|
||||
fn bound_record_duration(self, r: &'a BoundValueRecorder<f64>) -> BoxFuture<'a, Self::Output> {
|
||||
async move {
|
||||
let request_start = SystemTime::now();
|
||||
let res = self.await;
|
||||
r.record(request_start.elapsed().map_or(0.0, |d| d.as_secs_f64()));
|
||||
res
|
||||
}
|
||||
.boxed()
|
||||
}
|
||||
}
|
||||
|
||||
// ----
|
||||
|
||||
pub fn gen_trace_id() -> TraceId {
|
||||
rand::thread_rng().gen::<[u8; 16]>().into()
|
||||
}
|
||||
|
|
@ -0,0 +1,92 @@
|
|||
use std::sync::{
|
||||
atomic::{AtomicUsize, Ordering},
|
||||
Arc,
|
||||
};
|
||||
|
||||
use sled::{CompareAndSwapError, IVec, Iter, Result, Tree};
|
||||
|
||||
#[derive(Clone)]
|
||||
pub struct SledCountedTree(Arc<SledCountedTreeInternal>);
|
||||
|
||||
struct SledCountedTreeInternal {
|
||||
tree: Tree,
|
||||
len: AtomicUsize,
|
||||
}
|
||||
|
||||
impl SledCountedTree {
|
||||
pub fn new(tree: Tree) -> Self {
|
||||
let len = tree.len();
|
||||
Self(Arc::new(SledCountedTreeInternal {
|
||||
tree,
|
||||
len: AtomicUsize::new(len),
|
||||
}))
|
||||
}
|
||||
|
||||
pub fn len(&self) -> usize {
|
||||
self.0.len.load(Ordering::Relaxed)
|
||||
}
|
||||
|
||||
pub fn is_empty(&self) -> bool {
|
||||
self.0.tree.is_empty()
|
||||
}
|
||||
|
||||
pub fn get<K: AsRef<[u8]>>(&self, key: K) -> Result<Option<IVec>> {
|
||||
self.0.tree.get(key)
|
||||
}
|
||||
|
||||
pub fn iter(&self) -> Iter {
|
||||
self.0.tree.iter()
|
||||
}
|
||||
|
||||
// ---- writing functions ----
|
||||
|
||||
pub fn insert<K, V>(&self, key: K, value: V) -> Result<Option<IVec>>
|
||||
where
|
||||
K: AsRef<[u8]>,
|
||||
V: Into<IVec>,
|
||||
{
|
||||
let res = self.0.tree.insert(key, value);
|
||||
if res == Ok(None) {
|
||||
self.0.len.fetch_add(1, Ordering::Relaxed);
|
||||
}
|
||||
res
|
||||
}
|
||||
|
||||
pub fn pop_min(&self) -> Result<Option<(IVec, IVec)>> {
|
||||
let res = self.0.tree.pop_min();
|
||||
if let Ok(Some(_)) = &res {
|
||||
self.0.len.fetch_sub(1, Ordering::Relaxed);
|
||||
};
|
||||
res
|
||||
}
|
||||
|
||||
pub fn compare_and_swap<K, OV, NV>(
|
||||
&self,
|
||||
key: K,
|
||||
old: Option<OV>,
|
||||
new: Option<NV>,
|
||||
) -> Result<std::result::Result<(), CompareAndSwapError>>
|
||||
where
|
||||
K: AsRef<[u8]>,
|
||||
OV: AsRef<[u8]>,
|
||||
NV: Into<IVec>,
|
||||
{
|
||||
let old_some = old.is_some();
|
||||
let new_some = new.is_some();
|
||||
|
||||
let res = self.0.tree.compare_and_swap(key, old, new);
|
||||
|
||||
if res == Ok(Ok(())) {
|
||||
match (old_some, new_some) {
|
||||
(false, true) => {
|
||||
self.0.len.fetch_add(1, Ordering::Relaxed);
|
||||
}
|
||||
(true, false) => {
|
||||
self.0.len.fetch_sub(1, Ordering::Relaxed);
|
||||
}
|
||||
_ => (),
|
||||
}
|
||||
}
|
||||
res
|
||||
}
|
||||
}
|
||||
|
|
@ -1,6 +1,6 @@
|
|||
[package]
|
||||
name = "garage_web"
|
||||
version = "0.6.0"
|
||||
version = "0.7.0"
|
||||
authors = ["Alex Auvolat <alex@adnab.me>", "Quentin Dufour <quentin@dufour.io>"]
|
||||
edition = "2018"
|
||||
license = "AGPL-3.0"
|
||||
|
|
@ -14,16 +14,18 @@ path = "lib.rs"
|
|||
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
|
||||
|
||||
[dependencies]
|
||||
garage_api = { version = "0.6.0", path = "../api" }
|
||||
garage_model = { version = "0.6.0", path = "../model" }
|
||||
garage_util = { version = "0.6.0", path = "../util" }
|
||||
garage_table = { version = "0.6.0", path = "../table" }
|
||||
garage_api = { version = "0.7.0", path = "../api" }
|
||||
garage_model = { version = "0.7.0", path = "../model" }
|
||||
garage_util = { version = "0.7.0", path = "../util" }
|
||||
garage_table = { version = "0.7.0", path = "../table" }
|
||||
|
||||
err-derive = "0.3"
|
||||
log = "0.4"
|
||||
tracing = "0.1.30"
|
||||
percent-encoding = "2.1.0"
|
||||
|
||||
futures = "0.3"
|
||||
|
||||
http = "0.2"
|
||||
hyper = { version = "0.14", features = ["server", "http1", "runtime", "tcp", "stream"] }
|
||||
|
||||
opentelemetry = "0.17"
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
//! Crate for handling web serving of s3 bucket
|
||||
#[macro_use]
|
||||
extern crate log;
|
||||
extern crate tracing;
|
||||
|
||||
mod error;
|
||||
pub use error::Error;
|
||||
|
|
|
|||
|
|
@ -9,6 +9,13 @@ use hyper::{
|
|||
Body, Method, Request, Response, Server,
|
||||
};
|
||||
|
||||
use opentelemetry::{
|
||||
global,
|
||||
metrics::{Counter, ValueRecorder},
|
||||
trace::{FutureExt, TraceContextExt, Tracer},
|
||||
Context, KeyValue,
|
||||
};
|
||||
|
||||
use crate::error::*;
|
||||
|
||||
use garage_api::error::{Error as ApiError, OkOrBadRequest, OkOrInternalError};
|
||||
|
|
@ -20,6 +27,33 @@ use garage_model::garage::Garage;
|
|||
|
||||
use garage_table::*;
|
||||
use garage_util::error::Error as GarageError;
|
||||
use garage_util::metrics::{gen_trace_id, RecordDuration};
|
||||
|
||||
struct WebMetrics {
|
||||
request_counter: Counter<u64>,
|
||||
error_counter: Counter<u64>,
|
||||
request_duration: ValueRecorder<f64>,
|
||||
}
|
||||
|
||||
impl WebMetrics {
|
||||
fn new() -> Self {
|
||||
let meter = global::meter("garage/web");
|
||||
Self {
|
||||
request_counter: meter
|
||||
.u64_counter("web.request_counter")
|
||||
.with_description("Number of requests to the web endpoint")
|
||||
.init(),
|
||||
error_counter: meter
|
||||
.u64_counter("web.error_counter")
|
||||
.with_description("Number of requests to the web endpoint resulting in errors")
|
||||
.init(),
|
||||
request_duration: meter
|
||||
.f64_value_recorder("web.request_duration")
|
||||
.with_description("Duration of requests to the web endpoint")
|
||||
.init(),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// Run a web server
|
||||
pub async fn run_web_server(
|
||||
|
|
@ -28,13 +62,19 @@ pub async fn run_web_server(
|
|||
) -> Result<(), GarageError> {
|
||||
let addr = &garage.config.s3_web.bind_addr;
|
||||
|
||||
let metrics = Arc::new(WebMetrics::new());
|
||||
|
||||
let service = make_service_fn(|conn: &AddrStream| {
|
||||
let garage = garage.clone();
|
||||
let metrics = metrics.clone();
|
||||
|
||||
let client_addr = conn.remote_addr();
|
||||
async move {
|
||||
Ok::<_, Error>(service_fn(move |req: Request<Body>| {
|
||||
let garage = garage.clone();
|
||||
handle_request(garage, req, client_addr)
|
||||
let metrics = metrics.clone();
|
||||
|
||||
handle_request(garage, metrics, req, client_addr)
|
||||
}))
|
||||
}
|
||||
});
|
||||
|
|
@ -49,23 +89,55 @@ pub async fn run_web_server(
|
|||
|
||||
async fn handle_request(
|
||||
garage: Arc<Garage>,
|
||||
metrics: Arc<WebMetrics>,
|
||||
req: Request<Body>,
|
||||
addr: SocketAddr,
|
||||
) -> Result<Response<Body>, Infallible> {
|
||||
info!("{} {} {}", addr, req.method(), req.uri());
|
||||
match serve_file(garage, &req).await {
|
||||
|
||||
// Lots of instrumentation
|
||||
let tracer = opentelemetry::global::tracer("garage");
|
||||
let span = tracer
|
||||
.span_builder(format!("Web {} request", req.method()))
|
||||
.with_trace_id(gen_trace_id())
|
||||
.with_attributes(vec![
|
||||
KeyValue::new("method", format!("{}", req.method())),
|
||||
KeyValue::new("uri", req.uri().to_string()),
|
||||
])
|
||||
.start(&tracer);
|
||||
|
||||
let metrics_tags = &[KeyValue::new("method", req.method().to_string())];
|
||||
|
||||
// The actual handler
|
||||
let res = serve_file(garage, &req)
|
||||
.with_context(Context::current_with_span(span))
|
||||
.record_duration(&metrics.request_duration, &metrics_tags[..])
|
||||
.await;
|
||||
|
||||
// More instrumentation
|
||||
metrics.request_counter.add(1, &metrics_tags[..]);
|
||||
|
||||
// Returning the result
|
||||
match res {
|
||||
Ok(res) => {
|
||||
debug!("{} {} {}", req.method(), req.uri(), res.status());
|
||||
debug!("{} {} {}", req.method(), res.status(), req.uri());
|
||||
Ok(res)
|
||||
}
|
||||
Err(error) => {
|
||||
info!(
|
||||
"{} {} {} {}",
|
||||
req.method(),
|
||||
req.uri(),
|
||||
error.http_status_code(),
|
||||
req.uri(),
|
||||
error
|
||||
);
|
||||
metrics.error_counter.add(
|
||||
1,
|
||||
&[
|
||||
metrics_tags[0].clone(),
|
||||
KeyValue::new("status_code", error.http_status_code().to_string()),
|
||||
],
|
||||
);
|
||||
Ok(error_to_res(error))
|
||||
}
|
||||
}
|
||||
|
|
|
|||
Loading…
Reference in New Issue