REReplication · Interviews
Community homeHot discussions Latest posts Best of Knowledge base Resources About

Replication · Interviews · The edge cases the replication-interview docs never spell out

140
REr/replication-interview·posted by tang_hao·32 minutes agoOpen source

The edge cases the replication-interview docs never spell out

Short version: replication-interview 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.

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

Worth noting: the official docs do cover this, just in a very inconspicuous spot. I only found it reading the source comments, where the author explains the reasoning — roughly "so that it degrades into predictable behaviour in extreme cases".

Nnews.ycombinator.comExternal link · opens in a new tab
33 comments

33 comments

M
Aalice_dev·12 minutes ago

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

357
Zzhou_yiOP·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.

37
Ddev_zhou·2 days ago

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

467
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.

447
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.

66
Mmike_xu·28 minutes 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.

87
Zzhou_yi·just nowedited

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

420
Lli_ming·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.

420
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.

404
Ttang_hao·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.

54
Cchen_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.

402
Oops_wang·1 hour ago

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

3
Kkite·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.

1
Rrase·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.

6
Ddev_zhouMod·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.

365
Wwinter·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
Hhuang_ke·just now

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

187
Ttang_hao·28 minutes agoedited

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

175
Aalice_devOP·12 minutes ago

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

360
Rrase·5 hours agoedited

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

60
Ddev_zhou·2 hours ago

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

37
NnikicMod·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.

3
Wwinter·5 hours ago

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

105
RraseOP·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.

16
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.

36
Nnikic·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
Ttang_hao·28 minutes 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.

2
Sswoole_lee·12 minutes ago

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

49
Ddev_zhou·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.

370
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.

38
Bbob_chen·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.

30
Aalice_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.

2
Zzhou_yi·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/replication-interview/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 →