Comment 4 for bug 1867616
Revision history for this message
| Maciej Borzecki (maciek-borzecki) wrote Re: refresh of snapd failed, maybe because of concurrent refresh of core | #4 |
© 2004
Canonical Ltd.
•
Terms of use
•
Data privacy
•
Contact Launchpad Support
•
Blog
•
Careers
•
System status
•
bbfa235
(Get the code!)
From looking at the source code and the logs, I believe this is what happens:
- snapd snap is updated
- snapd requests a restart
- we reach the code in daemon.Stop() where we attempt a graceful shutdown of http connections
- the shutdown code reaches a default timeout (25s, matches with the log), while there's still active connections
- the listen socket is closed as the first step, so no new connections are possible
- the error of the tomb is set and propagated up the call stack
- snapd exits with an error triggering a failover handling which reverts the new revision
I suspect there may be a slow client hanging on the snapd API socket that keeps the connection alive. We should consider swallowing up the error like we do for system restarts, or using a forceful http.Server.Close() when the timeout is hit.