BABaking · Events
Community homeHot discussions Latest posts Best of Knowledge base Resources About

Baking · Events · baking-events under load on an 8-core 16GB box — full benchmark scripts included

1.1K
BAr/baking-events·posted by slow_query·6 hours agoTooling

baking-events under load on an 8-core 16GB box — full benchmark scripts included

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

Worth noting: the official docs do cover this, just in a very inconspicuous spot. I only found it reading the source comments, where the author explains the reasoning — roughly "so that it degrades into predictable behaviour in extreme cases".

On trade-offs, my view is this: if nobody on the team owns this area long-term, do not introduce a second mechanism. With two coexistence you first have to work out which one is even in play when things break, and that costs far more than the performance you saved.

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.

384 comments

384 comments

· first 120 loaded
M
Zzhou_yiOP·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.

409
Zzhou_yi·2 days ago

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

466
Llinlin·2 days ago

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

509
Bbob_chen·12 minutes agoedited

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

459
Kkite·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.

428
RraseOP·3 minutes ago

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

9
Sslow_query·2 hours agoedited

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.

6
Mmike_xu·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.

9
LlinlinOP·3 minutes ago

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

314
Sswoole_lee·12 minutes ago

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

347
Llinlin·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.

94
Hhuang_ke·28 minutes ago

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

286
Kkite·2 days ago

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

134
Kkernel_panic·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.

32
Cchen_dev·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.

128
Sslow_query·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.

172
Aalice_dev·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.

121
Kkite·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.

314
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.

24
Sswoole_leeMod·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.

335
Llinlin·2 days agoLevel 6

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

179
Sswoole_lee·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.

7
Bbob_chen·2 days ago

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

1
RraseOP·2 days ago

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

224
Bbob_chen·2 days ago

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

276
Hhuang_ke·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.

18
Cchen_dev·2 days ago

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

270
Lli_ming·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.

264
Rrase·2 days ago

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

261
Wwinter·2 days ago

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

260
Lli_ming·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.

254
Rran_bo·2 days agoedited

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.

96
Sswoole_lee·2 days ago

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

71
Aalice_dev·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.

395
Rran_bo·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.

15
Lli_ming·3 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.

253
Aalice_dev·2 days agoedited

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.

328
Cchen_dev·2 days agoedited

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

510
Ttang_hao·2 days ago

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

226
Mmike_xu·2 days ago

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

490
Zzhou_yi·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.

323
Kkernel_panic·2 days agoedited

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.

19
Sswoole_lee·2 days ago

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

138
Hhuang_ke·just nowedited

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

3
Bbob_chen·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.

232
Zzhou_yiMod·3 minutes agoedited

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.

17
Cchen_dev·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.

6
Wwinter·2 days agoedited

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

1
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.

482
Oops_wang·2 days ago

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

250
Sswoole_lee·5 hours ago

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

241
Rrase·2 days ago

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

189
Aalice_dev·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.

136
Mmike_xu·5 hours agoedited

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

152
Hhuang_ke·2 days 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.

148
Rran_bo·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.

126
Aalice_dev·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.

124
NnikicOP·2 days ago

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

478
Rran_bo·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.

118
Llinlin·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.

104
LlinlinOP·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.

39
Ttang_hao·2 hours ago

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

92
Wwinter·12 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.

91
Kkernel_panic·just now

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.

88
Zzhu_zong·28 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.

265
Wwinter·2 days ago

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

2
Rran_boOP·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.

22
Mmike_xu·2 days ago

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

82
Bbob_chenOP·2 days ago

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

18
Ttang_hao·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.

3
Sswoole_leeOP·2 days ago

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

114
Lli_ming·5 hours ago

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

80
Sslow_query·2 hours ago

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

75
Kkernel_panic·2 days ago

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

396
Hhuang_ke·2 days ago

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

1
Ttang_hao·2 days ago

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

9
Wwinter·2 days ago

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

65
Mmike_xu·2 days agoedited

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.

63
Ttang_hao·2 days ago

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

58
Zzhu_zong·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.

52
Rran_bo·5 hours ago

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

404
Ttang_hao·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.

425
Ddev_zhou·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.

395
Sswoole_lee·2 days agoedited

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.

220
Ddev_zhou·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
Sslow_query·2 days agoeditedLevel 6

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

1
Ddev_zhou·2 days 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.

212
Rrase·2 days ago

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

40
Zzhou_yi·2 days agoedited

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

39
Ttang_hao·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.

428
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.

37
Oops_wang·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.

34
Cchen_dev·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.

22
Zzhou_yi·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.

20
KkiteOP·2 days agoedited

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

448
KkiteMod·2 days ago

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

18
Llinlin·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.

17
Nnikic·2 days ago

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

54
Ddev_zhou·2 days ago

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

16
Rran_bo·2 days agoedited

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

397
Lli_ming·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.

14
Mmike_xu·5 hours 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.

10
Bbob_chen·2 days ago

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

10
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.

387
Cchen_dev·2 days ago

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

9
Sslow_query·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.

95
Wwinter·2 days ago

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

44
Kkite·just now

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

8
Sslow_query·just now

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.

60
Ttang_hao·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.

175
Mmike_xu·2 days 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.

5
Rrase·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.

4
Hhuang_ke·28 minutes 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.

3
Hhuang_ke·1 hour ago

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

3
Aalice_devMod·12 minutes ago

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

3
Llinlin·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.

3
Bbob_chen·2 days ago

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

297
Ttang_hao·12 minutes 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
Rran_bo·2 days ago

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

1
Wwinter·2 days ago

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

1

This is the post detail page /en/c/baking-events/post/p10. 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 →