SWSWC · Interviews
Community homeHot discussions Latest posts Best of Knowledge base Resources About

SWC · Interviews · Six rules for error handling in swc-interview that I settled on

700
SWr/swc-interview·posted by rase·32 minutes agoAnnouncement

Six rules for error handling in swc-interview that I settled on

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.

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.

What genuinely surprised me was the tail. The average looked great while P99 jumped by an order of magnitude past some threshold. The cause was not swc-interview itself but our upstream connection reuse — the load test traffic was too clean and hid the long-tail requests.

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.

313 comments

313 comments

· first 120 loaded
M
Rran_bo·2 days agoedited

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

513
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.

506
Lli_ming·2 days ago

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

16
Zzhu_zong·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.

446
Wwinter·2 days ago

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

386
Ttang_haoOPMod·2 days ago

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

319
Lli_ming·2 days ago

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

377
Sswoole_lee·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.

488
Zzhu_zong·12 minutes agoedited

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

356
Ttang_hao·3 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.

297
Rran_boOP·just now

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

65
Zzhu_zong·28 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.

384
Lli_ming·2 hours ago

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

520
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.

45
Kkernel_panic·2 days agoLevel 6

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

462
Lli_ming·2 days agoLevel 6

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

266
Zzhou_yi·2 hours 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.

431
Kkernel_panic·2 days ago

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

397
Sslow_query·2 days 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.

489
Rran_boOP·3 minutes agoLevel 6

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

50
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.

75
Wwinter·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.

1
Rrase·2 days agoLevel 6

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

1
Kkite·2 days ago

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

297
Mmike_xu·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.

257
Sslow_query·2 days agoLevel 6

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.

219
Mmike_xu·2 days agoLevel 6

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.

112
Rrase·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.

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.

111
Cchen_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.

384
Bbob_chen·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.

309
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.

69
Oops_wangOP·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.

190
Oops_wang·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.

59
Sslow_query·12 minutes 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.

80
Zzhu_zong·2 days agoLevel 6

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

30
Rrase·3 minutes ago

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

18
Oops_wangMod·12 minutes ago

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

293
Cchen_dev·2 days ago

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

273
Ddev_zhou·2 days ago

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

227
Lli_ming·just now

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

218
Wwinter·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.

11
Oops_wang·2 days ago

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

219
KkiteOP·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.

51
Lli_ming·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.

2
Kkernel_panic·2 days ago

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

103
Nnikic·2 days ago

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

393
Kkernel_panic·12 minutes ago

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

10
Rrase·2 days ago

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

65
Cchen_dev·12 minutes ago

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

211
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.

50
Hhuang_ke·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.

504
Bbob_chen·2 days ago

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

289
Bbob_chenOP·5 hours 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.

245
Zzhu_zong·2 days agoeditedLevel 6

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

17
Kkite·2 days agoeditedLevel 6

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

1
Ttang_hao·2 days ago

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

42
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.

35
Llinlin·2 days ago

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

262
Sslow_query·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.

222
Kkite·2 hours 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.

248
Kkernel_panicMod·2 days ago

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

1
Hhuang_ke·2 days agoLevel 6

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
Kkernel_panic·2 days agoLevel 6

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

99
Ddev_zhou·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.

7
Kkernel_panic·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.

1
Zzhu_zong·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.

494
Hhuang_ke·3 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.

80
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.

481
Rran_boOP·2 days ago

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

122
Kkite·2 days ago

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

1
Kkernel_panic·3 minutes ago

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

402
Sslow_query·2 hours 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.

41
Lli_ming·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.

292
Nnikic·1 hour ago

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

185
Rran_bo·2 days ago

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

205
Cchen_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.

189
Nnikic·2 days ago

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

148
Rran_boOP·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.

418
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.

147
Cchen_dev·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.

141
Kkernel_panic·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.

424
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.

1
Bbob_chen·yesterday

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

134
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.

384
Zzhu_zong·2 days ago

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

1
Rrase·yesterday

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

130
Bbob_chen·12 minutes ago

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

42
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.

123
Cchen_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.

107
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.

113
Nnikic·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.

95
Oops_wang·2 days agoedited

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

84
Rrase·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.

83
NnikicOP·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
WwinterOP·2 days ago

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

1
Ddev_zhou·2 days agoedited

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

54
Rran_bo·2 hours 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.

41
Zzhou_yi·12 minutes ago

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

91
Oops_wang·2 days ago

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

127
Mmike_xu·2 days ago

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

38
Oops_wang·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.

37
Ddev_zhou·1 hour ago

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

28
Ddev_zhou·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.

117
Ddev_zhou·2 days ago

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

26
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.

23
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.

17
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.

13
Kkite·2 days ago

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

13
Kkernel_panicMod·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.

13
Cchen_dev·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.

11
Cchen_dev·just now

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

14
Rran_bo·12 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.

302
Sswoole_lee·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.

1
Zzhou_yi·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.

2
Hhuang_ke·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.

4
Wwinter·2 days ago

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

4
Sslow_query·2 days ago

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

2
Zzhou_yi·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
Lli_ming·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/swc-interview/post/p4. 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 →