HCHigh Concurrency · Code review
Community homeHot discussions Latest posts Best of Knowledge base Resources About

High Concurrency · Code review · Open-sourced a rate-limiting middleware for high-concurrency-review with token bucket and sliding window

167
HCr/high-concurrency-review·posted by bob_chen·1 hour agoInternals

Open-sourced a rate-limiting middleware for high-concurrency-review with token bucket and sliding window

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.

-- 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;

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".

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.

48 comments

48 comments

M
Kkite·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.

325
Rran_bo·2 days ago

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

316
Cchen_devOP·12 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.

215
Aalice_dev·3 minutes ago

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

283
Ddev_zhou·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.

302
Sswoole_leeOP·just now

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.

75
Lli_ming·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.

230
Sswoole_lee·5 hours 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.

244
Rran_boOP·2 hours ago

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

265
Ddev_zhou·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.

226
Rrase·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.

13
Sslow_query·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.

48
Llinlin·28 minutes ago

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

202
Mmike_xu·2 days ago

One counter-example: below high-concurrency-review 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
Wwinter·28 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.

146
Kkite·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
LlinlinOP·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.

337
Aalice_devOP·just nowLevel 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.

377
Ttang_hao·2 days agoLevel 6

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

173
Cchen_dev·2 days agoLevel 6

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

6
Rrase·2 days ago

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

195
Bbob_chen·2 days 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.

172
Aalice_dev·2 days agoLevel 6

One counter-example: below high-concurrency-review 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
Nnikic·12 minutes 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.

254
Sslow_queryMod·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.

387
Bbob_chen·2 days agoedited

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

197
Kkernel_panic·3 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.

154
Hhuang_ke·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.

266
Cchen_dev·just now

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

472
Rran_bo·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.

251
Zzhu_zong·1 hour ago

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

454
Bbob_chen·5 hours ago

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

499
Ddev_zhou·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.

255
Sswoole_lee·2 days ago

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

2
Ddev_zhou·2 hours agoLevel 6

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.

162
Sslow_query·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.

190
Aalice_dev·2 hours agoedited

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

113
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.

90
Ttang_hao·2 days ago

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

322
KkiteMod·yesterday

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.

79
Zzhou_yi·12 minutes ago

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

77
Wwinter·2 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.

9
Cchen_devOP·3 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.

15
Kkite·2 days ago

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

219
Kkernel_panic·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.

25
Rrase·2 days ago

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

6
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.

12
Cchen_dev·2 hours ago

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

1

This is the post detail page /en/c/high-concurrency-review/post/p8. 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 →