Sometimes the lotus processes close the notifs channels and communication between processes fails. This leads to unpleasant to painful errors like failed wdPosts.
notifs channels seem not to be re established reliably after they closed.
- filecoin-project/lotus#8362
- filecoin-project/lotus#7653
- filecoin-project/lotus#6883
- filecoin-project/lotus#5813
At least:
- an entry for notifs channels in
lotus-miner log alerts
indicating that channels are OK/not OK
Better:
- an API endpoint to query the status of notifs channels
Ultimately:
- notifs channels being reliably reopened when closed or notifs channels not closing at all