UbUbuntu Community
Community homeReleases Service deployment System tuning

Ubuntu Community · Newbie question: how should a Ubuntu config file be organised?

547
Ubr/ubuntu·posted by zhu_zong·2 hours agoTranslation

Newbie question: how should a Ubuntu config file be organised?

Short version: Ubuntu 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.

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.

Documentation first44%
Source code first28%
Just ask someone17%
Run a demo and learn by error11%

930 votes total

92 comments

92 comments

M
Zzhou_yi·2 days ago

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

504
Hhuang_ke·3 minutes ago

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

492
Wwinter·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.

382
Aalice_devOP·3 minutes ago

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

272
Kkernel_panicOP·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.

89
Oops_wangOP·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.

1
Rrase·12 minutes ago

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

98
Cchen_dev·2 days ago

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

465
Zzhou_yi·12 minutes ago

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

80
Wwinter·2 days ago

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

232
Sswoole_lee·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.

5
RraseOP·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
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.

12
Aalice_dev·yesterday

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.

39
Zzhu_zong·2 days ago

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

255
Kkernel_panic·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.

220
Sslow_query·2 days ago

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

137
Kkernel_panic·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.

480
Rran_bo·2 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.

19
Cchen_dev·2 days agoedited

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

13
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.

1
Sswoole_leeMod·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.

91
Oops_wang·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.

308
Mmike_xuOP·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.

212
Sswoole_leeMod·just now

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

245
Rrase·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.

96
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.

23
Ddev_zhou·2 days ago

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

302
Kkite·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.

275
Rran_bo·2 days agoedited

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

272
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.

418
Kkernel_panic·2 days ago

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

354
Rrase·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.

216
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.

216
Aalice_dev·2 days ago

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

208
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.

204
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.

166
Mmike_xu·2 days ago

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

195
Oops_wang·2 days ago

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

170
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.

78
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.

45
Zzhu_zong·12 minutes ago

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

101
Hhuang_ke·2 days ago

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

94
Mmike_xuOP·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.

24
Ddev_zhou·2 days ago

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

82
Rrase·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.

81
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.

76
Sslow_queryOP·just now

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

1
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.

46
Mmike_xu·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.

1
Sswoole_lee·1 hour 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.

45
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.

43
Llinlin·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.

37
Sslow_query·2 days ago

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

7
Ttang_hao·1 hour ago

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

4
Zzhu_zong·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
Kkernel_panic·12 minutes ago

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

3
Ttang_hao·28 minutes agoedited

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

289
Sswoole_lee·yesterday

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

173
Rran_bo·2 days ago

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

133
Hhuang_ke·just now

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

152
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.

334
Lli_ming·yesterday

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.

210
Kkernel_panic·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.

182
Zzhu_zong·2 days ago

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

110
Hhuang_ke·2 days agoLevel 6

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

348
Rran_bo·2 days agoedited

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

106
Lli_ming·2 days ago

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

65
Kkernel_panic·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.

161
Zzhou_yi·2 days ago

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

477
Mmike_xu·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.

6
Bbob_chen·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
Oops_wang·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.

264
Lli_ming·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.

398
RraseMod·2 days agoeditedLevel 6

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

159
Oops_wang·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.

137
Rran_bo·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.

153
Zzhou_yi·28 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.

42
Llinlin·2 days ago

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

276
Zzhu_zong·2 days agoeditedLevel 6

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

133
Sslow_query·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.

20
Kkernel_panic·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
Nnikic·2 days agoedited

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

113
Mmike_xuMod·2 days ago

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

27
Wwinter·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.

1
Lli_ming·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.

11
Wwinter·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.

81
Wwinter·2 days ago

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

1
Ttang_hao·2 days ago

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

1
Rrase·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.

15
Ttang_hao·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

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

50

This is the post detail page /en/c/ubuntu/post/p9. 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 →