Alipay Mini Program Community · Two weeks with alipay-miniprogram: what feels right and what drives me up the wall

272
AMr/alipay-miniprogram·posted by linlin·32 minutes agoPostmortem

Two weeks with alipay-miniprogram: what feels right and what drives me up the wall

Most alipay-miniprogram articles stop at "how to use it" and never cover "when not to use it". This is an attempt at the second half.

# Load test: do not jump straight to the max concurrency.
# Ramp it up, otherwise you miss the knee.
for c in 50 100 200 400 800; do
  wrk -t8 -c$c -d60s --latency http://127.0.0.1:8080/api/feed
  sleep 20
done

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.

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.

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 alipay-miniprogram itself but our upstream connection reuse — the load test traffic was too clean and hid the long-tail requests.

115 comments

115 comments

M
WwinterOP·2 days ago

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

470
Ddev_zhou·2 days ago

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

514
Lli_ming·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.

499
Kkernel_panic·2 days ago

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

479
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.

491
Nnikic·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.

466
Ddev_zhou·5 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.

465
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.

434
Aalice_dev·yesterday

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

396
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.

100
Kkernel_panic·2 days ago

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

47
Zzhou_yi·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.

355
Mmike_xu·2 days ago

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

444
Bbob_chen·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.

26
Ttang_hao·2 days ago

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

23
Sslow_query·12 minutes ago

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

348
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.

343
Mmike_xu·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.

260
Sslow_query·28 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.

334
Nnikic·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.

91
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.

129
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.

122
Ddev_zhou·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.

31
Bbob_chenMod·2 days ago

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

428
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.

2
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.

302
Wwinter·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.

291
Rran_boOP·2 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.

230
Zzhou_yiOP·2 days ago

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

27
Ttang_hao·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.

281
Ttang_hao·28 minutes ago

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

493
Zzhu_zongMod·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.

66
Ttang_hao·1 hour agoedited

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

4
Nnikic·2 days ago

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

7
Sslow_query·28 minutes ago

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

459
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.

47
Llinlin·2 days agoedited

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

1
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.

406
Bbob_chen·12 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.

92
Cchen_dev·just now

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

226
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.

345
RraseOP·28 minutes ago

One counter-example: below alipay-miniprogram 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
Sswoole_lee·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.

50
Llinlin·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.

6
RraseOP·1 hour ago

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

17
Nnikic·just now

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

29
Aalice_dev·2 days ago

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

272
Sslow_queryOP·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.

195
Lli_ming·2 days ago

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

238
Nnikic·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.

211
Nnikic·2 days ago

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

191
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.

152
Mmike_xuOP·2 days ago

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

94
Sslow_query·2 days agoedited

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

223
Mmike_xu·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.

138
Zzhu_zong·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.

120
Sslow_query·12 minutes ago

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

372
Bbob_chenOP·2 days ago

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

41
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.

27
Llinlin·2 days ago

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

203
Llinlin·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.

47
Kkernel_panic·2 days ago

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

56
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.

392
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.

89
KkiteOP·2 days ago

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

211
Ddev_zhou·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.

1
Oops_wangOP·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.

2
Kkernel_panicOPMod·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.

31
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.

25
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.

1
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.

105
Bbob_chen·1 hour ago

One counter-example: below alipay-miniprogram 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
Rran_bo·2 days ago

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

2
Bbob_chen·28 minutes ago

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

97
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.

95
Kkernel_panic·2 days ago

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

95
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.

85
Aalice_dev·2 days agoedited

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

83
Ddev_zhou·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.

82
LlinlinOP·2 days agoedited

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

2
Rran_bo·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.

6
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.

15
Aalice_devOP·2 days ago

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

13
Wwinter·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.

67
WwinterMod·1 hour 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.

61
Ddev_zhou·2 days ago

One counter-example: below alipay-miniprogram 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
Bbob_chenOP·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.

1
Ddev_zhou·2 days ago

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

239
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.

1
Lli_ming·5 hours ago

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

6
Zzhou_yi·2 days agoedited

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

518
Aalice_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.

78
Zzhou_yi·2 days agoLevel 6

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

1
Sswoole_lee·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.

11
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.

55
Hhuang_keMod·2 days ago

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

55
Rran_bo·2 days ago

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

41
Bbob_chen·12 minutes ago

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

34
Kkernel_panic·2 days ago

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

30
Rrase·2 days ago

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

26
Zzhou_yiMod·2 days agoedited

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

24
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.

9
Ddev_zhou·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.

6
Kkite·2 days ago

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

247
Ddev_zhou·yesterday

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

6
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.

3
Mmike_xu·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.

2
Oops_wang·2 days ago

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

331
Kkite·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.

2
Oops_wang·2 days ago

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

2
Aalice_dev·yesterday

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

2
Zzhu_zong·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.

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
Lli_mingOP·2 days ago

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

149
Kkernel_panic·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/alipay-miniprogram/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 →