fix(bug): memory leak in tracing client#4828
Merged
SuperQ merged 3 commits intoprometheus:mainfrom Jan 9, 2026
Merged
Conversation
Signed-off-by: Cody Kaczynski <ckaczyns@akamai.com>
4742b52 to
84f13fa
Compare
ultrotter
reviewed
Jan 1, 2026
ultrotter
reviewed
Jan 1, 2026
Signed-off-by: Cody Kaczynski <ckaczyns@akamai.com>
fa55dc4 to
4c4f184
Compare
TheMeier
approved these changes
Jan 5, 2026
ultrotter
approved these changes
Jan 7, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
We (Akamai/Linode) upgraded our Alertmanager machines to v0.30.0 after some testing on December 18th, 2025 and started seeing the process being OOMKilled on December 23rd, 2025.
We investigated a bit and realized that there was a memory leak relating to the new distributed tracing feature, where each request instantiated a new Transport client.
Before
This is the heap from one of the machines we saw this on, running v0.30.0:
As you can see,
go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp.(*Transport).RoundTripis using ~50% of the heap.We found feat: add distributed tracing support was included in this release, and then noticed in
notify/util.gothat in therequestmethod, a new session was created for every request:To get around this, in this PR we're initializing a single client instead and just re-using it.
After
I'm testing it by spamming tens thousands of alerts, so memory allocation isn't necessarily what it is in a normal production workload, but anyways this is what the heap looks like after my change:
As you can see, much less! For what it's worth, we don't currently get traces from Alertmanager (yet), but it looks like my change didn't break anything:

Summary
We are now re-using a single Transport client for tracing instead of creating a new one for each request, resulting in a significant decrease in memory usage:
For what it's worth, the machine we tested on did not have any OOMKills within the last ~2-3 weeks that we've been testing v0.30.0. We only saw this in production, which are under a considerable amount more load.
I'm not sure what the urgency is here for other folks or if anyone else is seeing similar behavior, but we've rolled back to v0.28.1 for now so not too big of a deal for us (although we would like to get back up to v0.30.x soon!)
Signed-off-by: Cody Kaczynski ckaczyns@akamai.com