Skip to content

Recover Counterparty read slots after cancelled worker runs - #64

Merged
droplister merged 3 commits into
mainfrom
fix/expired-counterparty-read-slots
Oct 17, 2026
Merged

droplister merged 3 commits into
mainfrom
fix/expired-counterparty-read-slots

Conversation

@droplister

Copy link
Copy Markdown
Member

Scheduled indexing can stop indefinitely when a cancelled Worker invocation leaves both Counterparty read slots occupied. Later runs then spend the full 15-second timeout waiting for admission. Expire slots at their original read deadline, release them on abort, and make late cleanup harmless to replacement reads. Timeout logs now identify admission, fetch, or body processing.

Validated with a workerd reproduction of abandoned reads, 121 targeted tests, API type/lint checks, and production smoke checks. Deployed API version 93d043f3-c6a3-4dff-b109-c95682e5aec6: the index caught up from block 971882 to 971906, scheduled mempool reads returned to under one second, and Telegram announcements resumed at 20:56 UTC on October 17.

@droplister
droplister merged commit 5755022 into main Oct 17, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant