AMAlipay Mini Program Community
Community homeHot discussions Latest posts Best of Knowledge base Resources About

Alipay Mini Program Community · Why this alipay-miniprogram pattern gets slow in production: a source-level explanation

1.2K
AMr/alipay-miniprogram·posted by li_ming·2 hours agoTutorial

Why this alipay-miniprogram pattern gets slow in production: a source-level explanation

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;

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.

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.

293 comments

293 comments

· first 120 loaded
M
Wwinter·3 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.

512
Wwinter·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.

450
Rran_boMod·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.

405
Cchen_devOP·just now

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

327
Hhuang_ke·3 minutes 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.

377
Nnikic·2 days ago

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

373
Mmike_xu·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.

363
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.

346
WwinterOP·2 days ago

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

7
Cchen_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.

386
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.

24
Rrase·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.

282
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.

431
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.

362
Kkernel_panic·2 days ago

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

269
Zzhou_yi·just now

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".

258
Nnikic·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.

257
Kkernel_panic·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.

256
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.

43
Hhuang_ke·12 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.

213
Lli_ming·2 hours 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.

403
Rran_bo·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.

205
NnikicOP·just now

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

91
Hhuang_keOP·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.

25
Ddev_zhou·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.

149
Kkite·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.

463
Bbob_chen·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.

302
Rran_bo·2 days agoLevel 6

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

31
Mmike_xu·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.

12
Hhuang_ke·2 days agoeditedLevel 6

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

320
Zzhou_yiOP·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.

108
Mmike_xu·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.

69
Nnikic·2 days ago

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

47
Kkernel_panic·28 minutes ago

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

84
Sslow_query·2 days ago

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

210
Nnikic·2 days ago

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

260
Wwinter·2 hours ago

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

127
LlinlinOP·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.

4
Zzhou_yi·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.

140
Lli_ming·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.

93
Ddev_zhou·28 minutes ago

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

79
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.

344
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.

1
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.

174
Cchen_dev·2 days agoLevel 6

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.

9
Kkernel_panic·1 hour 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.

158
Mmike_xu·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.

183
Zzhou_yi·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.

107
Lli_ming·28 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.

30
Mmike_xu·2 days agoLevel 6

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.

393
Sswoole_lee·28 minutes agoeditedLevel 6

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

274
Kkernel_panic·yesterdayLevel 6

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.

264
Llinlin·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.

241
Aalice_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.

25
Kkite·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.

55
Sswoole_lee·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.

1
Zzhu_zong·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.

5
Oops_wang·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.

2
Kkite·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".

1
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".

65
Hhuang_ke·2 days agoLevel 6

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".

41
Oops_wangOP·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.

76
Ttang_hao·2 days ago

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

130
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.

127
Lli_ming·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.

411
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".

111
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.

107
Kkernel_panic·yesterdayedited

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.

122
RraseOP·2 days ago

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

30
Oops_wang·5 hours agoedited

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

183
Sswoole_lee·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.

78
Sslow_queryOP·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.

57
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.

2
Kkernel_panic·2 days ago

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

73
Rran_bo·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.

260
Bbob_chenOP·28 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.

13
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.

71
Zzhu_zongOP·2 days ago

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

7
Rrase·1 hour ago

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

63
Lli_ming·2 days ago

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

60
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.

50
Llinlin·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.

49
Bbob_chen·2 days ago

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

12
Lli_ming·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".

329
Mmike_xuMod·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.

69
Ddev_zhou·2 days agoedited

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

12
Aalice_dev·12 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.

48
Rran_bo·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.

201
Zzhu_zong·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.

1
Aalice_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.

48
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.

44
Bbob_chen·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.

25
Llinlin·2 days agoedited

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

356
Hhuang_ke·just now

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

13
Aalice_dev·12 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.

69
Cchen_dev·3 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.

450
Kkernel_panic·yesterday

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

389
Cchen_dev·2 days agoedited

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

11
Rrase·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.

21
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.

146
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.

1
Zzhu_zong·2 days agoedited

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

12
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.

11
Hhuang_ke·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".

10
Hhuang_ke·yesterday

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.

185
Zzhu_zong·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.

42
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.

30
Bbob_chen·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.

3
Rrase·2 days ago

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

8
Hhuang_ke·3 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".

446
Cchen_dev·2 days ago

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

7
Wwinter·2 days ago

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

5
Bbob_chen·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.

5
Lli_ming·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
Ttang_hao·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.

1
Sswoole_lee·2 days ago

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

95
Zzhu_zong·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.

3
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.

509
Rrase·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
Zzhou_yi·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.

1

This is the post detail page /en/c/alipay-miniprogram/post/p6. 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 →