TETensorRT · Beginners
Community homeHot discussions Latest posts Best of Knowledge base Resources About

TensorRT · Beginners · Ten tensorrt-newbie anti-patterns — how many are you guilty of?

69
TEr/tensorrt-newbie·posted by alice_dev·5 minutes agoOpen sourceLocked

Ten tensorrt-newbie anti-patterns — how many are you guilty of?

Most tensorrt-newbie articles stop at "how to use it" and never cover "when not to use it". This is an attempt at the second half.

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

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 tensorrt-newbie 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".

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.

21 comments

21 comments

M
Ddev_zhou·2 days agoedited

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

498
Oops_wang·3 minutes ago

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

224
Llinlin·3 minutes 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.

13
Aalice_devOP·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.

158
Oops_wang·28 minutes ago

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

193
Ttang_hao·5 hours ago

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

93
Lli_ming·12 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.

163
Llinlin·1 hour ago

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

92
Mmike_xuMod·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.

301
Ttang_haoOP·2 days ago

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

47
LlinlinOP·5 hours ago

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

26
Hhuang_ke·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.

76
Ddev_zhou·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.

17
Cchen_dev·5 hours agoedited

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

69
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.

4
Nnikic·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.

35
Ttang_hao·3 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.

32
Zzhu_zong·28 minutes ago

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

25
Lli_ming·28 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.

18
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.

6
LlinlinOP·2 days ago

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

30

This is the post detail page /en/c/tensorrt-newbie/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 →