Wasmtime GitHub notifications bot (Jul 29 2022 at 19:47): Wasmtime GitHub notifications bot (Jul 29 2022 at 19:47):peterhuene commented on issue #4556:
OOM kill on the arm64 linux test?
Wasmtime GitHub notifications bot (Jul 29 2022 at 19:59):alexcrichton commented on issue #4556:
That's probably because some new test running is making a large virtual mapping and it's causing qemu to actually reserve memory and OOM. I'm not sure which test though so I'll take a look next week.
Wasmtime GitHub notifications bot (Jul 29 2022 at 20:18):github-actions[bot] commented on issue #4556:
Subscribe to Label Action
cc @peterhuene
<details>
This issue or pull request has been labeled: "wasmtime:api"Thus the following users have been cc'd because of the following labels:
- peterhuene: wasmtime:api
To subscribe or unsubscribe from this label, edit the <code>.github/subscribe-to-label.json</code> configuration file.
Learn more.
</details>
Wasmtime GitHub notifications bot (Aug 01 2022 at 15:14):alexcrichton commented on issue #4556:
The best I can surmise is that when we allocate virtual memory QEMU keeps some sort of flags data structure behind the scenes which means that a massive virtual memory reservation allocates memory on the host proportional to the size. This is nothing new since we already disable all pooling allocator tests within QEMU since the TB-sized reservation consumes too much memory on the host.
My current guess is that the component model tests create lots of instances which creates lots of 4G reservations for linear memories (which I guess aren't reused? unsure) so I've tried disabling "static" memory and instead forcing dynamic memory for all component model tests when run under qemu. Will see how CI responds...
Last updated: Dec 23 2024 at 12:05 UTC