rune/garage
1
0
Fork 0
forked from Deuxfleurs/garage
Code Issues Pull requests Projects Releases Wiki Activity

[next-0.10] cluster node status metrics: report nodes of all active layout versions

Browse source
This commit is contained in:
Alex 2024-02-20 17:08:31 +01:00
parent 643d1aabd8
commit 75e591727d
Signed by untrusted user: lx
GPG key ID: 0E496D15096376BE
1 changed files with 34 additions and 34 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

30
src/rpc/system_metrics.rs
View file

@ -216,12 +216,12 @@ impl SystemMetrics {
.u64_value_observer("cluster_layout_node_connected", move |observer| { .u64_value_observer("cluster_layout_node_connected", move |observer| {
let layout = system.cluster_layout(); let layout = system.cluster_layout();
let nodes = system.get_known_nodes(); let nodes = system.get_known_nodes();
for (id, _, config) in layout.current().roles.items().iter() { for id in layout.all_nodes().iter() {
if let Some(role) = &config.0 { let mut kv = vec![KeyValue::new("id", format!("{:?}", id))];
let mut kv = vec![ if let Some(role) =
KeyValue::new("id", format!("{:?}", id)), layout.current().roles.get(id).and_then(|r| r.0.as_ref())
KeyValue::new("role_zone", role.zone.clone()), {
]; kv.push(KeyValue::new("role_zone", role.zone.clone()));
match role.capacity { match role.capacity {
Some(cap) => { Some(cap) => {
kv.push(KeyValue::new("role_capacity", cap as i64)); kv.push(KeyValue::new("role_capacity", cap as i64));
@ -231,10 +231,10 @@ impl SystemMetrics {
kv.push(KeyValue::new("role_gateway", 1)); kv.push(KeyValue::new("role_gateway", 1));
} }
} }
}
let value; let value;
if let Some(node) = nodes.iter().find(|n| n.id == *id) { if let Some(node) = nodes.iter().find(|n| n.id == *id) {
value = if node.is_up { 1 } else { 0 };
// TODO: if we add address and hostname, and those change, we // TODO: if we add address and hostname, and those change, we
// get duplicate metrics, due to bad otel aggregation :( // get duplicate metrics, due to bad otel aggregation :(
// Can probably be fixed when we upgrade opentelemetry // Can probably be fixed when we upgrade opentelemetry
@ -243,13 +243,13 @@ impl SystemMetrics {
// "hostname", // "hostname",
// node.status.hostname.clone(), // node.status.hostname.clone(),
// )); // ));
value = if node.is_up { 1 } else { 0 };
} else { } else {
value = 0; value = 0;
} }
observer.observe(value, &kv); observer.observe(value, &kv);
} }
}
}) })
.with_description("Connection status for nodes in the cluster layout") .with_description("Connection status for nodes in the cluster layout")
.init() .init()
@ -260,12 +260,12 @@ impl SystemMetrics {
.u64_value_observer("cluster_layout_node_disconnected_time", move |observer| { .u64_value_observer("cluster_layout_node_disconnected_time", move |observer| {
let layout = system.cluster_layout(); let layout = system.cluster_layout();
let nodes = system.get_known_nodes(); let nodes = system.get_known_nodes();
for (id, _, config) in layout.current().roles.items().iter() { for id in layout.all_nodes().iter() {
if let Some(role) = &config.0 { let mut kv = vec![KeyValue::new("id", format!("{:?}", id))];
let mut kv = vec![ if let Some(role) =
KeyValue::new("id", format!("{:?}", id)), layout.current().roles.get(id).and_then(|r| r.0.as_ref())
KeyValue::new("role_zone", role.zone.clone()), {
]; kv.push(KeyValue::new("role_zone", role.zone.clone()));
match role.capacity { match role.capacity {
Some(cap) => { Some(cap) => {
kv.push(KeyValue::new("role_capacity", cap as i64)); kv.push(KeyValue::new("role_capacity", cap as i64));
@ -275,6 +275,7 @@ impl SystemMetrics {
kv.push(KeyValue::new("role_gateway", 1)); kv.push(KeyValue::new("role_gateway", 1));
} }
} }
}
if let Some(node) = nodes.iter().find(|n| n.id == *id) { if let Some(node) = nodes.iter().find(|n| n.id == *id) {
// TODO: see comment above // TODO: see comment above
@ -290,7 +291,6 @@ impl SystemMetrics {
} }
} }
} }
}
}) })
.with_description( .with_description(
"Time (in seconds) since last connection to nodes in the cluster layout", "Time (in seconds) since last connection to nodes in the cluster layout",