NINixOS · Toolchain
Community homeHot discussions Latest posts Best of Knowledge base Resources About

NixOS · Toolchain · Help: nixos-tools fails to start intermittently in containers, logs inside

179
NIr/nixos-tools·posted by kernel_panic·1 hour agoDiscussion

Help: nixos-tools fails to start intermittently in containers, logs inside

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.

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

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

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.

73 comments

73 comments

M
Aalice_devOP·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.

450
Ddev_zhou·2 days ago

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

126
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.

488
Sswoole_lee·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.

421
Wwinter·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.

231
Bbob_chenOP·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.

126
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.

178
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.

166
Zzhou_yi·2 days ago

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

161
Wwinter·just now

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

103
Llinlin·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.

58
Lli_ming·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.

294
Cchen_dev·1 hour agoedited

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

1
Lli_mingOP·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.

162
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.

1
Sswoole_lee·2 days ago

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

104
Sslow_query·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.

181
Ttang_haoOP·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.

7
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.

269
Aalice_dev·just now

One counter-example: below nixos-tools 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
Ttang_hao·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.

174
Nnikic·2 days ago

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

6
Zzhu_zongOP·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
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.

8
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
Sslow_query·just nowedited

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

2
Kkite·2 hours ago

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

97
RraseOP·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.

126
Oops_wangOP·2 days ago

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

411
Nnikic·2 days ago

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

1
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.

17
Aalice_devOP·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.

303
Zzhu_zong·2 days ago

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

221
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.

15
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.

11
Rrase·2 days ago

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

46
Ttang_hao·1 hour ago

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

43
Sslow_query·2 hours ago

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

29
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.

28
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.

19
Zzhu_zong·2 days ago

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

18
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.

17
Wwinter·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.

159
Hhuang_ke·12 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.

15
Rrase·just nowedited

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.

66
Cchen_devMod·2 days ago

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

6
Zzhou_yi·28 minutes ago

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

393
Kkernel_panic·2 days ago

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

1
Aalice_dev·2 days ago

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

311
Oops_wang·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.

142
Ttang_hao·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.

7
Hhuang_ke·2 days agoedited

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

80
Bbob_chen·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.

1
Sswoole_leeMod·2 days agoLevel 6

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

77
Oops_wang·2 days ago

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

9
Bbob_chen·1 hour ago

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

7
Wwinter·2 hours ago

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

416
Wwinter·2 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.

1
Oops_wang·3 minutes ago

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

3
Kkernel_panic·1 hour ago

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

2
Nnikic·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.

1
Hhuang_ke·2 days ago

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

27
Oops_wang·2 days agoedited

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

515
Oops_wang·2 days ago

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

284
Rran_bo·2 days ago

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

1
Lli_ming·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.

53
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.

1
WwinterOP·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.

156
Zzhou_yi·2 days ago

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

34
Nnikic·2 days ago

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

127
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.

18
Hhuang_ke·2 days agoedited

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

1
Kkernel_panic·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.

1

This is the post detail page /en/c/nixos-tools/post/p3. 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 →