ALAlertmanager · English
Community homeHot discussions Latest posts Best of Knowledge base Resources About

Alertmanager · English · Those easily-missed type details in alertmanager-en

129
ALr/alertmanager-en·posted by winter·3 hours agoTutorial

Those easily-missed type details in alertmanager-en

Some background first. Our setup is alertmanager-en plus three downstream services, seven figures of daily requests, peaking around nine in the evening.

-- The query that broke: a full scan over 20M rows.
-- A composite index took P99 from 1.8s down to 42ms.
SELECT id, title, created_at
  FROM posts
 WHERE community_id = ?
   AND status = 1
 ORDER BY score DESC
 LIMIT 20;

One last trap: in container environments remember to adjust the memory-related parameters in step. Otherwise the host limit and the process expectation disagree, and the symptom is intermittent, unreproducible failure.

We also fixed monitoring along the way: replaced average-based alerts with percentiles and split them per endpoint. False alerts dropped by about seventy percent and the on-call rotation visibly cheered up.

The first thing was to collapse the variables. We were changing config and upgrading the version at the same time, and afterwards nobody could say which change caused what. We rolled back to moving one variable at a time, re-ran three times, and only then did the curve settle. Tedious, but not skippable.

51 comments

51 comments

M
Oops_wangOP·3 minutes agoedited

Sharing our numbers, 8 cores 16GB, same scenario:

| Concurrency | P50 | P99 |
|---|---|---|
| 200 | 12ms | 88ms |
| 500 | 31ms | 340ms |

P99 clearly collapses at 500 concurrency, which lines up with your knee point.

418
Oops_wangOP·5 hours ago

This is not a alertmanager-en problem, it is a usage problem. The docs say this API is not thread-safe and you must lock around it yourself.

493
Cchen_dev·1 hour ago

Can you give a minimal reproduction? I ran it locally for ten minutes and could not reproduce on macOS with the latest version.

37
Zzhou_yi·2 days ago

I just read the alertmanager-en source — the author actually explains the reasoning in a comment, roughly "so that it degrades into predictable behaviour in extreme cases".

444
Nnikic·3 minutes ago

Saved. I am reworking this area this week — this saves a lot of wrong turns.

135
Nnikic·2 days ago

This matches what we see in production. We only hit it past 3k QPS; the earlier load tests showed nothing — the test traffic was too clean, with no long-tail requests.

33
Cchen_devMod·2 days ago

I see point 3 differently. The trade-off depends on your read/write ratio: read-heavy with little writing means caching actually widens the inconsistency window.

17
NnikicMod·1 hour ago

There is actually a simpler fix that needs no architecture change: move this check up to the gateway and the problem disappears. The cost is one extra lookup at the gateway.

59
Sslow_query·3 minutes ago

I just read the alertmanager-en source — the author actually explains the reasoning in a comment, roughly "so that it degrades into predictable behaviour in extreme cases".

144
Sswoole_lee·2 days agoLevel 6

Agreeing with the above. One addition: with this option enabled the GC count in your metrics doubles, so adjust the alert threshold at the same time or it will keep firing.

219
Kkite·just now

Thanks for sharing real numbers — far more useful than the articles that only cover concepts.

57
Ddev_zhouOP·2 days ago

Sharing our numbers, 8 cores 16GB, same scenario:

| Concurrency | P50 | P99 |
|---|---|---|
| 200 | 12ms | 88ms |
| 500 | 31ms | 340ms |

P99 clearly collapses at 500 concurrency, which lines up with your knee point.

264
Lli_ming·1 hour ago

Has anyone run a controlled experiment? I did, reducing it to a single variable, and the difference was 4% — within noise. So I suspect the main cause is something else.

293
Zzhu_zong·2 days agoedited

A question: what changes in a container with a 512Mi memory limit? That is how we run it in production.

206
Hhuang_ke·2 days ago

Has anyone run a controlled experiment? I did, reducing it to a single variable, and the difference was 4% — within noise. So I suspect the main cause is something else.

21
Rrase·2 days ago

There is actually a simpler fix that needs no architecture change: move this check up to the gateway and the problem disappears. The cost is one extra lookup at the gateway.

213
Kkernel_panic·5 hours ago

Worth learning from this debugging approach. We went straight at the logs and took a much longer route.

64
Hhuang_ke·2 days agoedited

Sharing our numbers, 8 cores 16GB, same scenario:

| Concurrency | P50 | P99 |
|---|---|---|
| 200 | 12ms | 88ms |
| 500 | 31ms | 340ms |

P99 clearly collapses at 500 concurrency, which lines up with your knee point.

219
Sslow_query·12 minutes ago

I see point 3 differently. The trade-off depends on your read/write ratio: read-heavy with little writing means caching actually widens the inconsistency window.

2
Lli_ming·2 days ago

This is not a alertmanager-en problem, it is a usage problem. The docs say this API is not thread-safe and you must lock around it yourself.

430
Wwinter·just nowedited

This matches what we see in production. We only hit it past 3k QPS; the earlier load tests showed nothing — the test traffic was too clean, with no long-tail requests.

307
Bbob_chen·2 days ago

Agreeing with the above. One addition: with this option enabled the GC count in your metrics doubles, so adjust the alert threshold at the same time or it will keep firing.

30
Ddev_zhou·2 days ago

This matches what we see in production. We only hit it past 3k QPS; the earlier load tests showed nothing — the test traffic was too clean, with no long-tail requests.

138
Mmike_xu·2 days ago

Thanks for sharing real numbers — far more useful than the articles that only cover concepts.

279
Kkite·2 days ago

This is not a alertmanager-en problem, it is a usage problem. The docs say this API is not thread-safe and you must lock around it yourself.

190
Kkernel_panic·2 days ago

Sharing our numbers, 8 cores 16GB, same scenario:

| Concurrency | P50 | P99 |
|---|---|---|
| 200 | 12ms | 88ms |
| 500 | 31ms | 340ms |

P99 clearly collapses at 500 concurrency, which lines up with your knee point.

25
Sslow_query·28 minutes ago

A question: what changes in a container with a 512Mi memory limit? That is how we run it in production.

260
KkiteOP·2 days ago

Saved. I am reworking this area this week — this saves a lot of wrong turns.

292
Bbob_chen·2 days ago

Worth learning from this debugging approach. We went straight at the logs and took a much longer route.

193
Sslow_query·28 minutes ago

One counter-example: below alertmanager-en 7.4 the semantics of that code are different, so do not copy it verbatim. We got burned in staging and rolled back once.

89
Mmike_xu·2 days agoedited

We have run this in production for two years without hitting it. That said, we never reached this scale, so our experience is not really evidence here.

490
Sslow_query·2 days ago

Has anyone run a controlled experiment? I did, reducing it to a single variable, and the difference was 4% — within noise. So I suspect the main cause is something else.

188
Ddev_zhou·3 minutes agoedited

Thanks for sharing real numbers — far more useful than the articles that only cover concepts.

172
Lli_ming·yesterday

Has anyone run a controlled experiment? I did, reducing it to a single variable, and the difference was 4% — within noise. So I suspect the main cause is something else.

125
Rrase·2 days ago

Can you give a minimal reproduction? I ran it locally for ten minutes and could not reproduce on macOS with the latest version.

86
Oops_wang·1 hour ago

Saved. I am reworking this area this week — this saves a lot of wrong turns.

5
Hhuang_ke·28 minutes ago

Worth learning from this debugging approach. We went straight at the logs and took a much longer route.

33
Rran_boOP·28 minutes agoedited

We have run this in production for two years without hitting it. That said, we never reached this scale, so our experience is not really evidence here.

356
Kkite·12 minutes ago

Saved. I am reworking this area this week — this saves a lot of wrong turns.

26
Ddev_zhou·28 minutes ago

This matches what we see in production. We only hit it past 3k QPS; the earlier load tests showed nothing — the test traffic was too clean, with no long-tail requests.

12
Bbob_chen·3 minutes ago

One counter-example: below alertmanager-en 7.4 the semantics of that code are different, so do not copy it verbatim. We got burned in staging and rolled back once.

481
Oops_wangOPMod·2 days ago

One counter-example: below alertmanager-en 7.4 the semantics of that code are different, so do not copy it verbatim. We got burned in staging and rolled back once.

362
Bbob_chen·yesterday

There is actually a simpler fix that needs no architecture change: move this check up to the gateway and the problem disappears. The cost is one extra lookup at the gateway.

267
Hhuang_keOP·2 days ago

We have run this in production for two years without hitting it. That said, we never reached this scale, so our experience is not really evidence here.

6
Rran_bo·2 days ago

I just read the alertmanager-en source — the author actually explains the reasoning in a comment, roughly "so that it degrades into predictable behaviour in extreme cases".

36
Nnikic·1 hour ago

A question: what changes in a container with a 512Mi memory limit? That is how we run it in production.

7
Sslow_query·2 days ago

We have run this in production for two years without hitting it. That said, we never reached this scale, so our experience is not really evidence here.

4
Hhuang_ke·1 hour ago

I see point 3 differently. The trade-off depends on your read/write ratio: read-heavy with little writing means caching actually widens the inconsistency window.

235
Llinlin·2 hours agoedited

Can you give a minimal reproduction? I ran it locally for ten minutes and could not reproduce on macOS with the latest version.

26
Llinlin·12 minutes agoedited

Thanks for sharing real numbers — far more useful than the articles that only cover concepts.

9
Rrase·just now

Agreeing with the above. One addition: with this option enabled the GC count in your metrics doubles, so adjust the alert threshold at the same time or it will keep firing.

1

This is the post detail page /en/c/alertmanager-en/post/p5. Posts and comments are generated deterministically from a seeded PRNG, so the same post always renders the same content and the link can be shared, reloaded and indexed. In production this page reads MySQL for the post, Redis for hot-post caching, and fetches the whole comment tree in a single query on the path column.

See the database schema →