REReact · Architecture
Community homeHot discussions Latest posts Best of Knowledge base Resources About

React · Architecture · Six rules for error handling in react-design that I settled on

829
REr/react-design·posted by bob_chen·just nowAnnouncement

Six rules for error handling in react-design that I settled on

Short version: react-design needs almost no tuning at small and medium scale — the point where it starts to hurt is much further out than most people assume. Full measurements below.

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.

Performance44%
Maintainability28%
Ecosystem and community17%
Hiring difficulty11%

1409 votes total

166 comments

166 comments

· first 120 loaded
M
Oops_wang·2 days ago

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

472
Wwinter·2 days ago

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

418
Llinlin·2 days 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.

343
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.

139
Oops_wang·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.

402
Nnikic·5 hours ago

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

389
Aalice_dev·1 hour 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.

385
Lli_ming·1 hour ago

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

108
Zzhou_yi·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.

43
Sswoole_lee·2 days ago

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

332
Kkite·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.

325
Nnikic·2 days agoedited

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

351
Mmike_xu·yesterday

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.

300
Ttang_hao·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.

287
Ddev_zhouOP·2 days ago

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

207
Kkernel_panicOP·1 hour ago

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

199
KkiteMod·12 minutes ago

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

225
Ddev_zhou·2 hours ago

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

301
Oops_wangOP·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.

416
Wwinter·28 minutes ago

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

294
Lli_ming·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.

179
Llinlin·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.

102
Llinlin·2 days agoLevel 6

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

54
Zzhou_yi·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.

47
Oops_wang·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.

4
Lli_ming·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
Aalice_devMod·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.

1
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.

287
Oops_wang·3 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.

383
Llinlin·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.

308
Cchen_dev·2 days ago

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

90
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.

273
Sslow_query·2 days ago

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

498
Zzhou_yi·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.

7
Wwinter·1 hour ago

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

295
Lli_ming·2 days agoLevel 6

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

271
Sswoole_leeOP·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.

5
Mmike_xu·28 minutes 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.

23
Aalice_devMod·2 days agoLevel 6

One counter-example: below react-design 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
Lli_mingOP·2 days ago

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

11
Sswoole_lee·2 days agoLevel 6

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

304
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.

2
Ddev_zhouOP·2 days agoLevel 6

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

511
Kkernel_panicOP·2 days agoLevel 6

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

321
Aalice_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.

240
Hhuang_ke·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.

193
Bbob_chen·2 days ago

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

114
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.

82
Zzhu_zong·2 days agoLevel 6

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

462
Cchen_dev·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.

216
Hhuang_ke·2 days agoLevel 6

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

7
Mmike_xu·3 minutes ago

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

258
Kkernel_panicOP·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.

203
Zzhu_zong·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.

318
Sslow_query·1 hour ago

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

185
Wwinter·2 days agoLevel 6

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

1
Zzhu_zong·2 days ago

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

128
Hhuang_keMod·2 days ago

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

89
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.

27
Kkernel_panic·5 hours 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.

141
Sswoole_lee·28 minutes ago

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

235
Sswoole_lee·2 days ago

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

56
Mmike_xu·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.

431
Wwinter·28 minutes ago

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

94
Sslow_query·2 days agoLevel 6

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

1
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.

55
Wwinter·2 days ago

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

15
Ttang_hao·2 days ago

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

473
Wwinter·2 days ago

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

1
Oops_wang·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.

18
Mmike_xu·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.

234
Oops_wang·2 days ago

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

257
Kkernel_panicOP·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.

157
Sswoole_leeOP·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.

139
LlinlinOP·2 days ago

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

128
Mmike_xu·2 days agoedited

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

180
Cchen_devOP·2 days ago

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

112
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.

343
Sslow_query·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.

275
Hhuang_ke·2 days ago

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

157
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.

93
Sslow_query·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.

175
Kkite·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.

126
Nnikic·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.

395
Kkite·yesterday

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

141
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.

241
Zzhou_yi·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.

5
Llinlin·2 days ago

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

81
Cchen_dev·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.

87
Zzhou_yiOP·2 days agoedited

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

2
Sswoole_lee·28 minutes ago

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

79
Llinlin·2 days ago

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

68
Ddev_zhouOP·2 days ago

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

1
Kkernel_panicOP·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
Cchen_dev·28 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.

42
Llinlin·2 days ago

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

165
Zzhu_zong·5 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.

101
Llinlin·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.

68
Mmike_xu·yesterday

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

1
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.

232
Hhuang_ke·2 days agoedited

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

211
Ddev_zhouOPMod·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.

37
Zzhu_zong·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.

41
Cchen_dev·2 days ago

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

38
Rrase·2 days ago

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

17
Llinlin·2 days ago

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

16
Kkite·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.

284
Zzhou_yi·2 days ago

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

14
Kkite·2 days agoedited

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

13
Kkernel_panic·2 hours agoedited

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

12
Nnikic·2 days ago

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

12
Zzhou_yi·2 days ago

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

10
Sswoole_lee·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.

5
Cchen_dev·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.

74
Wwinter·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.

68
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.

2
Rrase·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.

334
Rrase·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.

1
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
Sswoole_lee·2 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.

1

This is the post detail page /en/c/react-design/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 →