storageAPI restart behaviour

Not planned

Follow

Kris Shannon

November 08, 2018 23:11

We had a recent issue during the upgrade from 5.5 to 6.0 where the storageAPI service on a large majority of our KVM hypervisors didn't restart properly.  They all failed with the error: 

socket.error: No socket could be created

This was probably due to a previous child process still holding on to the *:8080 socket.

I would like to suggest changing the systemd unit configuration.

Either changing `KillMode` from `process` to `mixed` which should make sure there are no lingering children from the previous run.

Or changing `Restart` from `on-abort` to `on-failure` (or even `always`) which may have allowed recovery - assuming the old child process didn't take too long to exit.

0

• Official comment

  

  Roman Strazhnyk (Product Manager) January 04, 2019 11:56

  Hi,

  Our development team has reviewed your request but we weren't able to reproduce the issue. It looks like the storage services were started for the second time, this might be a result of some failures during the upgrade procedure. I would recommend to create a support request in case you see this issue again so that our support engineers can investigate this in more detail.

  Comment actions Permalink

Post is closed for comments.