Skip to content

Fix/artifact manager auto reconnect#86

Merged
cccoolll merged 2 commits into
mainfrom
fix/artifact-manager-auto-reconnect
Apr 26, 2026
Merged

Fix/artifact manager auto reconnect#86
cccoolll merged 2 commits into
mainfrom
fix/artifact-manager-auto-reconnect

Conversation

@cccoolll

Copy link
Copy Markdown
Collaborator

No description provided.

cccoolll and others added 2 commits April 25, 2026 16:32
When the Artifact Manager becomes unreachable, the snapshot upload
proxy goes stale.  Previously there was no recovery short of a service
restart.  Now SquidArtifactManager.refresh_service() re-acquires the
proxy from the existing WebSocket connection, and the upload path
automatically retries once with the fresh proxy on timeout/connection
errors.

Also rate-limits flush-loop warnings to once per 60 seconds and logs
exception type names so asyncio.TimeoutError no longer appears blank.

Co-Authored-By: DeepSeek-V4-Pro <noreply@deepseek.com>
The reconnect logic in _upload_to_artifact_manager only caught
TimeoutError/ConnectionError/OSError, but hypha_rpc RPC failures
raise RemoteException.  During a brief Artifact Manager outage at
21:28 UTC, 4 snaps failed with RemoteException without triggering
a proxy refresh.  The images were still saved locally and recovered
by the flush loop, but adding RemoteException to the retry trigger
will recover faster next time.

Co-Authored-By: DeepSeek-V4-Pro <noreply@deepseek.com>
@cccoolll
cccoolll merged commit eedf64d into main Apr 26, 2026
4 checks 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