adjust docs for replication factor

This commit is contained in:
Yureka 2024-03-04 18:42:17 +01:00
parent 603604cdfc
commit 8f86af52ed
5 changed files with 83 additions and 62 deletions

View file

@ -116,7 +116,7 @@ metadata_dir = "/var/lib/garage/meta"
data_dir = "/var/lib/garage/data" data_dir = "/var/lib/garage/data"
db_engine = "lmdb" db_engine = "lmdb"
replication_mode = "3" replication_factor = 3
compression_level = 2 compression_level = 2

View file

@ -12,7 +12,7 @@ An introduction to building cluster layouts can be found in the [production depl
In Garage, all of the data that can be stored in a given cluster is divided In Garage, all of the data that can be stored in a given cluster is divided
into slices which we call *partitions*. Each partition is stored by into slices which we call *partitions*. Each partition is stored by
one or several nodes in the cluster one or several nodes in the cluster
(see [`replication_mode`](@/documentation/reference-manual/configuration.md#replication_mode)). (see [`replication_factor`](@/documentation/reference-manual/configuration.md#replication_factor)).
The layout determines the correspondence between these partitions, The layout determines the correspondence between these partitions,
which exist on a logical level, and actual storage nodes. which exist on a logical level, and actual storage nodes.

View file

@ -59,7 +59,7 @@ metadata_dir = "/tmp/meta"
data_dir = "/tmp/data" data_dir = "/tmp/data"
db_engine = "lmdb" db_engine = "lmdb"
replication_mode = "none" replication_factor = 1
rpc_bind_addr = "[::]:3901" rpc_bind_addr = "[::]:3901"
rpc_public_addr = "127.0.0.1:3901" rpc_public_addr = "127.0.0.1:3901"

View file

@ -8,7 +8,8 @@ weight = 20
Here is an example `garage.toml` configuration file that illustrates all of the possible options: Here is an example `garage.toml` configuration file that illustrates all of the possible options:
```toml ```toml
replication_mode = "3" replication_factor = 3
consistency_mode = "consistent"
metadata_dir = "/var/lib/garage/meta" metadata_dir = "/var/lib/garage/meta"
data_dir = "/var/lib/garage/data" data_dir = "/var/lib/garage/data"
@ -90,7 +91,8 @@ Top-level configuration options:
[`lmdb_map_size`](#lmdb_map_size), [`lmdb_map_size`](#lmdb_map_size),
[`metadata_dir`](#metadata_dir), [`metadata_dir`](#metadata_dir),
[`metadata_fsync`](#metadata_fsync), [`metadata_fsync`](#metadata_fsync),
[`replication_mode`](#replication_mode), [`replication_factor`](#replication_factor),
[`consistency_mode`](#consistency_mode),
[`rpc_bind_addr`](#rpc_bind_addr), [`rpc_bind_addr`](#rpc_bind_addr),
[`rpc_bind_outgoing`](#rpc_bind_outgoing), [`rpc_bind_outgoing`](#rpc_bind_outgoing),
[`rpc_public_addr`](#rpc_public_addr), [`rpc_public_addr`](#rpc_public_addr),
@ -133,11 +135,12 @@ The `[admin]` section:
### Top-level configuration options ### Top-level configuration options
#### `replication_mode` {#replication_mode} #### `replication_factor` {#replication_factor}
Garage supports the following replication modes: The replication factor can be any positive integer smaller or equal the node count in your cluster.
The chosen replication factor has a big impact on the cluster's failure tolerancy and performance characteristics.
- `none` or `1`: data stored on Garage is stored on a single node. There is no - `1`: data stored on Garage is stored on a single node. There is no
redundancy, and data will be unavailable as soon as one node fails or its redundancy, and data will be unavailable as soon as one node fails or its
network is disconnected. Do not use this for anything else than test network is disconnected. Do not use this for anything else than test
deployments. deployments.
@ -148,17 +151,6 @@ Garage supports the following replication modes:
before losing data. Data remains available in read-only mode when one node is before losing data. Data remains available in read-only mode when one node is
down, but write operations will fail. down, but write operations will fail.
- `2-dangerous`: a variant of mode `2`, where written objects are written to
the second replica asynchronously. This means that Garage will return `200
OK` to a PutObject request before the second copy is fully written (or even
before it even starts being written). This means that data can more easily
be lost if the node crashes before a second copy can be completed. This
also means that written objects might not be visible immediately in read
operations. In other words, this mode severely breaks the consistency and
durability guarantees of standard Garage cluster operation. Benefits of
this mode: you can still write to your cluster when one node is
unavailable.
- `3`: data stored on Garage will be stored on three different nodes, if - `3`: data stored on Garage will be stored on three different nodes, if
possible each in a different zones. Garage tolerates two node failure, or possible each in a different zones. Garage tolerates two node failure, or
several node failures but in no more than two zones (in a deployment with at several node failures but in no more than two zones (in a deployment with at
@ -166,55 +158,84 @@ Garage supports the following replication modes:
or node failures are only in a single zone, reading and writing data to or node failures are only in a single zone, reading and writing data to
Garage can continue normally. Garage can continue normally.
- `3-degraded`: a variant of replication mode `3`, that lowers the read - `5`, `7`, ...: When setting the replication factor above 3, it is most useful to
quorum to `1`, to allow you to read data from your cluster when several choose an uneven value, since for every two copies added, one more node can fail
nodes (or nodes in several zones) are unavailable. In this mode, Garage before losing the ability to write and read to the cluster.
does not provide read-after-write consistency anymore. The write quorum is
still 2, ensuring that data successfully written to Garage is stored on at
least two nodes.
- `3-dangerous`: a variant of replication mode `3` that lowers both the read
and write quorums to `1`, to allow you to both read and write to your
cluster when several nodes (or nodes in several zones) are unavailable. It
is the least consistent mode of operation proposed by Garage, and also one
that should probably never be used.
Note that in modes `2` and `3`, Note that in modes `2` and `3`,
if at least the same number of zones are available, an arbitrary number of failures in if at least the same number of zones are available, an arbitrary number of failures in
any given zone is tolerated as copies of data will be spread over several zones. any given zone is tolerated as copies of data will be spread over several zones.
**Make sure `replication_mode` is the same in the configuration files of all nodes. **Make sure `replication_factor` is the same in the configuration files of all nodes.
Never run a Garage cluster where that is not the case.** Never run a Garage cluster where that is not the case.**
It is technically possible to change the replication factor although it's a
dangerous operation that is not officially supported. This requires you to
delete the existing cluster layout and create a new layout from scratch,
meaning that a full rebalancing of your cluster's data will be needed. To do
it, shut down your cluster entirely, delete the `custer_layout` files in the
meta directories of all your nodes, update all your configuration files with
the new `replication_factor` parameter, restart your cluster, and then create a
new layout with all the nodes you want to keep. Rebalancing data will take
some time, and data might temporarily appear unavailable to your users.
It is recommended to shut down public access to the cluster while rebalancing
is in progress. In theory, no data should be lost as rebalancing is a
routine operation for Garage, although we cannot guarantee you that everything
will go right in such an extreme scenario.
#### `consistency_mode` {#consistency_mode}
The consistency mode setting determines the read and write behaviour of your cluster.
- `consistent`: The default setting. This is what the paragraph above describes.
The read and write quorum will be determined so that read-after-write consistency
is guaranteed.
- `degraded`: Lowers the read
quorum to `1`, to allow you to read data from your cluster when several
nodes (or nodes in several zones) are unavailable. In this mode, Garage
does not provide read-after-write consistency anymore.
The write quorum stays the same as in the `consistent` mode, ensuring that
data successfully written to Garage is stored on multiple nodes (depending
the replication factor).
- `dangerous`: This mode lowers both the read
and write quorums to `1`, to allow you to both read and write to your
cluster when several nodes (or nodes in several zones) are unavailable. It
is the least consistent mode of operation proposed by Garage, and also one
that should probably never be used.
Changing the `consistency_mode` between modes while leaving the `replication_factor` untouched
(e.g. setting your node's `consistency_mode` to `degraded` when it was previously unset, or from
`dangerous` to `consistent`), can be done easily by just changing the `consistency_mode`
parameter in your config files and restarting all your Garage nodes.
The consistency mode can be used together with various replication factors, to achieve
a wide range of read and write characteristics. Some examples:
- Replication factor `2`, consistency mode `degraded`: While this mode
technically exists, its properties are the same as with consistency mode `consistent`,
since the read quorum with replication factor `2`, consistency mode `consistent` is already 1.
- Replication factor `2`, consistency mode `dangerous`: written objects are written to
the second replica asynchronously. This means that Garage will return `200
OK` to a PutObject request before the second copy is fully written (or even
before it even starts being written). This means that data can more easily
be lost if the node crashes before a second copy can be completed. This
also means that written objects might not be visible immediately in read
operations. In other words, this configuration severely breaks the consistency and
durability guarantees of standard Garage cluster operation. Benefits of
this configuration: you can still write to your cluster when one node is
unavailable.
The quorums associated with each replication mode are described below: The quorums associated with each replication mode are described below:
| `replication_mode` | Number of replicas | Write quorum | Read quorum | Read-after-write consistency? | | `consistency_mode` | `replication_factor` | Write quorum | Read quorum | Read-after-write consistency? |
| ------------------ | ------------------ | ------------ | ----------- | ----------------------------- | | ------------------ | -------------------- | ------------ | ----------- | ----------------------------- |
| `none` or `1` | 1 | 1 | 1 | yes | | `consistent` | 1 | 1 | 1 | yes |
| `2` | 2 | 2 | 1 | yes | | `consistent` | 2 | 2 | 1 | yes |
| `2-dangerous` | 2 | 1 | 1 | NO | | `dangerous` | 2 | 1 | 1 | NO |
| `3` | 3 | 2 | 2 | yes | | `consistent` | 3 | 2 | 2 | yes |
| `3-degraded` | 3 | 2 | 1 | NO | | `degraded` | 3 | 2 | 1 | NO |
| `3-dangerous` | 3 | 1 | 1 | NO | | `dangerous` | 3 | 1 | 1 | NO |
Changing the `replication_mode` between modes with the same number of replicas
(e.g. from `3` to `3-degraded`, or from `2-dangerous` to `2`), can be done easily by
just changing the `replication_mode` parameter in your config files and restarting all your
Garage nodes.
It is also technically possible to change the replication mode to a mode with a
different numbers of replicas, although it's a dangerous operation that is not
officially supported. This requires you to delete the existing cluster layout
and create a new layout from scratch, meaning that a full rebalancing of your
cluster's data will be needed. To do it, shut down your cluster entirely,
delete the `custer_layout` files in the meta directories of all your nodes,
update all your configuration files with the new `replication_mode` parameter,
restart your cluster, and then create a new layout with all the nodes you want
to keep. Rebalancing data will take some time, and data might temporarily
appear unavailable to your users. It is recommended to shut down public access
to the cluster while rebalancing is in progress. In theory, no data should be
lost as rebalancing is a routine operation for Garage, although we cannot
guarantee you that everything will go right in such an extreme scenario.
#### `metadata_dir` {#metadata_dir} #### `metadata_dir` {#metadata_dir}

View file

@ -39,10 +39,10 @@ Read about cluster layout management [here](@/documentation/operations/layout.md
### Several replication modes ### Several replication modes
Garage supports a variety of replication modes, with 1 copy, 2 copies or 3 copies of your data, Garage supports a variety of replication modes, with configurable replica count,
and with various levels of consistency, in order to adapt to a variety of usage scenarios. and with various levels of consistency, in order to adapt to a variety of usage scenarios.
Read our reference page on [supported replication modes](@/documentation/reference-manual/configuration.md#replication_mode) Read our reference page on [supported replication modes](@/documentation/reference-manual/configuration.md#replication_factor)
to select the replication mode best suited to your use case (hint: in most cases, `replication_mode = "3"` is what you want). to select the replication mode best suited to your use case (hint: in most cases, `replication_factor = 3` is what you want).
### Compression and deduplication ### Compression and deduplication