Playground: cover the second class of TAP-busy leak (post-teardown zombies) - #1207
Merged
Merged
Conversation
…mbies)
Root cause of the recurring
FirecrackerError('PUT /snapshot/load -> 400: Resource busy (os error 16).
Invalid TUN/TAP Backend provided by fc-tap-<slot>')
was in _teardown, not _restore_snapshot. Evidence: 20 orphaned fc
processes on the box, all with 200 MiB..16 GiB RSS — i.e. VMs that
successfully restored, ran, and were later kicked (idle-reaper /
ensure-failed-retry / post-query-error). Something in _shutdown +
_teardown wasn't finishing them off. Two fixes here:
1. _shutdown: `asyncio.wait_for(vm.proc.wait(), timeout=5.0)` is too
short. Post-SIGKILL cleanup of a 16 GiB guest can take many
seconds (page freeing + virtio-blk fd close + memory-backing
unmap), especially under concurrent shutdowns. When wait_for
timed out, vm.proc/vm.pid were reset but the OS process was
still alive holding fc-tap-<slot>. Bump the reap window to 60 s
and follow up with an explicit _pid_alive poll that re-SIGKILLs
on each iteration. Log an error if the process outlives that
window (previously silent).
2. _teardown: also call net.teardown_tap unconditionally. Even in
the (now-rare) case a lingering fc still holds the TAP fd, the
next /restore's ensure_tap re-creates a fresh TAP. Cheap
(single tuntap add + addr), guaranteed clean.
Fixes the "many systems fail with Resource busy fc-tap-<N>" reports
across chdb-dataframe, clickhouse, starrocks, polars-dataframe,
doris-parquet, heavyai, spark-comet, victorialogs, quickwit,
trino-partitioned, clickhouse-web, and others.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Root cause of the recurring
was in _teardown, not _restore_snapshot. Evidence: 20 orphaned fc processes on the box, all with 200 MiB..16 GiB RSS — i.e. VMs that successfully restored, ran, and were later kicked (idle-reaper / ensure-failed-retry / post-query-error). Something in _shutdown + _teardown wasn't finishing them off. Two fixes here:
_shutdown:
asyncio.wait_for(vm.proc.wait(), timeout=5.0)is too short. Post-SIGKILL cleanup of a 16 GiB guest can take many seconds (page freeing + virtio-blk fd close + memory-backing unmap), especially under concurrent shutdowns. When wait_for timed out, vm.proc/vm.pid were reset but the OS process was still alive holding fc-tap-. Bump the reap window to 60 s and follow up with an explicit _pid_alive poll that re-SIGKILLs on each iteration. Log an error if the process outlives that window (previously silent)._teardown: also call net.teardown_tap unconditionally. Even in the (now-rare) case a lingering fc still holds the TAP fd, the next /restore's ensure_tap re-creates a fresh TAP. Cheap (single tuntap add + addr), guaranteed clean.
Fixes the "many systems fail with Resource busy fc-tap-" reports across chdb-dataframe, clickhouse, starrocks, polars-dataframe, doris-parquet, heavyai, spark-comet, victorialogs, quickwit, trino-partitioned, clickhouse-web, and others.