Skip to content

chore: add label to nvlink metric #521

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
tmcroberts97 wants to merge 1 commit into from
Closed

chore: add label to nvlink metric #521

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 8 additions & 4 deletions crates/api/src/nvl_partition_monitor/metrics.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,8 +50,9 @@ pub struct NvlPartitionMonitorMetrics {
pub num_stale_partitions_deleted: usize,
pub applied_changes: HashMap<AppliedChange, usize>,
pub operation_latencies: HashMap<AppliedChange, Vec<Duration>>,
/// Time from nvlink_config_version for instances currently in Pending (time spent in Pending), in milliseconds
pub nvlink_config_apply_durations_ms: Vec<f64>,
/// Time since instance config was requested
/// Each entry is (duration_ms, instance_id) for per-instance metrics.
pub nvlink_config_apply_durations_ms: Vec<(f64, String)>,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
Expand Down Expand Up @@ -336,8 +337,11 @@ impl NvlPartitionMonitorInstruments {
}
}

for &duration_ms in &metrics.nvlink_config_apply_durations_ms {
self.nvlink_config_apply_latency.record(duration_ms, &[]);
for (duration_ms, instance_id) in &metrics.nvlink_config_apply_durations_ms {
self.nvlink_config_apply_latency.record(
*duration_ms,
&[KeyValue::new("instance_id", instance_id.clone())],
Copy link
Contributor

@Matthias247 Matthias247 Mar 11, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this will lead to a lot of time-series. e.g. in a site with 1000 instances, you would get 1000 * amount_of_histogram_buckets new lines in the prometheus metrics file.

I'd avoid it, and think about how we can communicate high-latency issues in a different fashion.

);
}
}

Expand Down
4 changes: 3 additions & 1 deletion crates/api/src/nvl_partition_monitor/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1318,7 +1318,9 @@ impl NvlPartitionMonitor {
let duration_ms = (Utc::now() - instance.nvlink_config_version.timestamp())
.num_milliseconds()
.max(0) as f64;
metrics.nvlink_config_apply_durations_ms.push(duration_ms);
metrics
.nvlink_config_apply_durations_ms
.push((duration_ms, instance.id.to_string()));
}
}
}
Expand Down
Loading