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

Alipay Mini Program Community · After reading the alipay-miniprogram core source I finally understand the trade-offs

881
AMr/alipay-miniprogram·posted by slow_query·3 hours agoExperience

After reading the alipay-miniprogram core source I finally understand the trade-offs

Some background first. Our setup is alipay-miniprogram plus three downstream services, seven figures of daily requests, peaking around nine in the evening.

Order of investigation, by return on effort: 1. Check downstream latency first — usually it is not your problem 2. Then pool hit rate and wait-queue length 3. Only then GC and allocation 4. Suspect the framework last

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 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.

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.

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.

212 comments

212 comments

· first 120 loaded
M
Oops_wang·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.

445
Ddev_zhou·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.

425
Llinlin·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.

423
Sswoole_lee·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.

413
Nnikic·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.

47
Lli_ming·2 days ago

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

341
Rran_bo·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.

407
Hhuang_ke·yesterday

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.

4
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.

3
Mmike_xu·12 minutes ago

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

130
Zzhou_yi·2 days ago

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

429
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.

2
Hhuang_ke·28 minutes ago

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

387
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.

277
Sswoole_lee·2 hours 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.

373
Bbob_chen·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
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.

1
Wwinter·2 days ago

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

7
Sswoole_lee·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.

19
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.

470
Nnikic·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
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.

47
WwinterOP·12 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".

319
Wwinter·2 days ago

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

347
RraseOP·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.

374
Oops_wang·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.

2
Ttang_hao·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.

340
Kkernel_panic·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.

331
Zzhu_zong·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.

303
Mmike_xu·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.

367
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.

377
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.

117
Bbob_chen·5 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.

104
Cchen_dev·5 hours ago

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

132
KkiteOP·2 days ago

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

151
Mmike_xu·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".

22
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.

141
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.

283
Llinlin·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.

101
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.

3
Llinlin·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.

270
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.

263
Cchen_dev·3 minutes ago

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

251
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.

234
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.

310
Nnikic·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".

292
Ddev_zhou·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.

99
Nnikic·28 minutes ago

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

73
Lli_ming·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.

138
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.

15
Zzhu_zong·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.

119
Bbob_chen·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.

3
Ddev_zhou·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.

2
Lli_ming·5 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.

1
LlinlinOP·1 hour ago

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

147
Zzhou_yi·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·28 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.

236
Kkite·just now

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.

226
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.

198
Kkite·5 hours 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".

194
Rran_bo·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.

194
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.

192
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.

118
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.

78
Ddev_zhou·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.

309
LlinlinOP·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".

129
Llinlin·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.

172
Nnikic·2 days ago

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

1
Aalice_devOP·yesterday

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

77
Hhuang_ke·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.

246
Oops_wang·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.

129
Sswoole_lee·2 days ago

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

129
Zzhou_yi·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.

211
Rran_bo·2 days ago

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

118
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.

88
Kkernel_panicOPMod·2 days ago

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

22
Zzhou_yiOP·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".

21
Sswoole_lee·just now

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.

77
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.

65
Ttang_hao·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.

63
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.

62
LlinlinOP·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.

3
Zzhou_yiOP·5 hours 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.

1
Kkernel_panic·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.

7
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.

60
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.

77
Mmike_xu·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.

40
Lli_ming·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".

35
Aalice_dev·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.

28
Lli_ming·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.

24
Rrase·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.

20
Sslow_query·5 hours agoedited

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

15
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.

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.

14
Ddev_zhou·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.

9
Ddev_zhou·28 minutes 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.

12
Ttang_hao·5 hours 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.

319
Mmike_xu·just now

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.

317
Kkite·5 hours ago

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

19
Rrase·just now

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.

8
Oops_wang·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.

65
Ttang_hao·2 days agoLevel 6

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

68
Aalice_devOP·2 days agoLevel 6

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

1
Mmike_xuMod·12 minutes ago

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

10
Mmike_xu·2 days ago

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

8
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.

43
Bbob_chen·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".

6
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.

6
Sslow_query·2 days ago

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

452
Cchen_dev·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
Lli_ming·2 days ago

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

5
Lli_ming·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.

5
Oops_wang·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.

4
Zzhu_zong·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.

6
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.

32
Bbob_chen·yesterday

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

3
Zzhou_yi·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.

46
Sslow_query·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
Cchen_devMod·2 days ago

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

1
Sswoole_lee·2 days ago

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

1

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