Hello Flow Access Node Operators,
Recently an issue was identified where Access Nodes would hang during shutdown and only exit when force-killed by the container supervisor: Access node hangs when shutting down · Issue #8666 · onflow/flow-go · GitHub. The root cause was that long-lived streaming subscriptions (e.g. block/event streams) would keep grpc.Server.GracefulStop() blocked indefinitely, preventing the AN process from exiting cleanly.
A fix has been implemented and validated end-to-end on the foundation testnet and mainnet Access Nodes:
- PR: [v0.51] [AN] Backport graceful shutdown fix (#8662, #8664) by vishalchangrani · Pull Request #8676 · onflow/flow-go · GitHub
- Tag: Release v0.51.1: Graceful shutdown fix for Access Node (issue #8666) · onflow/flow-go · GitHub
Who this affects: Access Node operators only. Consensus, Collection, Execution, and Verification nodes are unaffected.
This is not urgent. Please update at your convenience.
Docker images available:
-
v0.51.1 -
v0.51.1-without-adx -
v0.51.1-without-netgo-without-adx -
v0.51.1-arm
Important notes:
- No changes to the flags
- Do NOT delete your data folder
- No coordination with other operators required (this is not a height-coordinated upgrade)
- For operators running multiple Access Nodes, you can stagger updates however you like there’s no strict ordering required.