fix(server): evict disk KV entry that fails prefill#253
Open
gmontana wants to merge 1 commit into
Open
Conversation
After an unclean shutdown the on-disk KV checkpoint can be intact (header, hash, token count all valid) but leave Metal in a state where prefill fails. Since the file keeps passing load-time checks it gets reloaded on every request, looping forever until the user manually deletes the cache directory. On prefill failure, if the prefix came from a disk entry, unlink it and invalidate the session. Next request gets a clean cache miss. Closes antirez#251
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes the infinite prefill failure loop from #251.
When ds4-server is killed mid-session, the disk KV checkpoint survives
and passes all load-time checks (header, hash, token count) on restart.
But the restored Metal state can be unusable — prefill fails, the error
goes back to the client, client retries, same checkpoint gets loaded
again, same failure. Loop runs forever until you rm -rf the cache dir.
The file is fine structurally, so the loader can't catch this. The fix:
when prefill fails and the prefix came from a disk entry, unlink that
file and invalidate the session. Next request misses the deleted entry
and does a clean prefill. Same pattern as the existing corrupt-prefix
discard path in the KV loader.
One helper (
kv_cache_evict_failed_disk_entry), two call sites ingenerate_job(), +13/-1.Tested by injecting a sync failure after a disk cache load — confirmed
the eviction log fires, the file disappears, and the next request
recovers with a full prefill. The actual Metal Internal Error from the
report is hardware-specific and not reproducible in CI.
Closes #251