RERedis · Ops
Community homeHot discussions Latest posts Best of Knowledge base Resources About

Redis · Ops · Help: redis-ops fails to start intermittently in containers, logs inside

371
REr/redis-ops·posted by alice_dev·32 minutes agoTooling

Help: redis-ops fails to start intermittently in containers, logs inside

It took me two weeks of on-and-off digging and plenty of wrong turns. Writing the process down as it happened so the next person spends less time.

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.

Image placeholder · object storage in production
POST /api/uploads → CDN origin pull
66 comments

66 comments

M
Nnikic·28 minutes ago

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

517
Ttang_hao·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.

175
Hhuang_ke·2 days ago

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

189
Rran_bo·1 hour ago

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

134
RraseMod·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.

32
Kkernel_panic·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.

41
Cchen_dev·2 hours ago

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

82
Rrase·2 days ago

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

23
Nnikic·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.

1
Nnikic·28 minutes ago

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

512
Sslow_query·2 days ago

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

422
Sswoole_lee·2 days agoedited

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

287
Hhuang_ke·3 minutes ago

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

196
Oops_wangMod·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.

85
Sswoole_lee·2 days ago

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

38
Wwinter·3 minutes 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.

244
Bbob_chenOP·28 minutes agoedited

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

338
Kkernel_panic·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.

108
Sslow_queryOP·2 days ago

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

180
Mmike_xu·2 days ago

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

223
Oops_wang·1 hour ago

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

135
Lli_ming·just now

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

406
Ddev_zhou·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.

39
Oops_wang·2 days ago

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

255
WwinterOP·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.

27
Sswoole_lee·yesterday

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

471
Kkite·2 days ago

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

54
Aalice_dev·28 minutes 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.

41
Ttang_hao·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.

41
Kkite·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.

34
Oops_wang·2 days ago

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

453
Aalice_dev·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.

10
Oops_wang·2 days ago

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

115
Ddev_zhou·2 days ago

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

28
Rrase·5 hours 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.

25
Zzhou_yi·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.

265
Hhuang_ke·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.

72
Sslow_query·2 days ago

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

25
Zzhu_zong·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.

23
Cchen_dev·12 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.

20
Ttang_hao·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.

8
Mmike_xu·12 minutes ago

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

7
Bbob_chen·2 days ago

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

2
Oops_wang·12 minutes agoedited

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.

1
Bbob_chen·just now

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.

490
Sswoole_lee·just nowedited

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

493
Hhuang_ke·3 minutes 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.

398
Nnikic·2 hours 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.

1
Nnikic·1 hour 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.

463
Aalice_dev·yesterdayedited

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.

209
Ddev_zhou·just now

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

85
Aalice_dev·3 minutes ago

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

37
Kkernel_panic·2 days agoLevel 6

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.

335
Mmike_xu·5 hours agoLevel 6

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.

239
Zzhu_zong·yesterday

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

244
Rrase·2 days ago

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

182
Bbob_chen·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.

2
Oops_wangOP·2 days agoLevel 6

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

2
Cchen_dev·2 hours agoeditedLevel 6

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

28
Ttang_hao·5 hours agoeditedLevel 6

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

21
Rrase·12 minutes ago

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

1
Rrase·yesterday

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.

462
Sslow_query·12 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.

3
Kkernel_panicMod·28 minutes ago

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

1
Bbob_chen·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.

1
Kkite·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.

1

This is the post detail page /en/c/redis-ops/post/p0. 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 →