#3990 server install failure during client enrollment shouldn't roll back
Closed: Fixed None Opened 8 years ago by rcritten.

Endi had a problem installing his IPA server. It kept failing during the client enrollment, which when failing tries to roll back the changes made. I think that in master mode the rollback should be skipped so you end up with a more debugable result.

2013-10-21T18:34:09Z DEBUG Could not connect socket to, error: (PR_CONNECT_RESET_ERROR) TCP connection reset by peer.
2013-10-21T18:34:09Z DEBUG Try to continue with next family...
2013-10-21T18:34:09Z ERROR Cannot connect to the server due to generic error: cannot connect to 'https://cloud-qe-4.idm.lab.bos.redhat.com/ipa/xml': Could not connect to cloud-qe-4.idm.lab.bos.redhat.com using any address: (PR_ADDRESS_NOT_SUPPORTED_ERROR) Network address type not supported.
2013-10-21T18:34:09Z ERROR Installation failed. Rolling back changes.

As a side note, it looks like the installation was failing because Apache wasn't quite starting up properly due to exhausting all the system semaphores. This was evidenced by the error message:

[Mon Oct 21 14:53:08.684987 2013] [core:emerg] [pid 490] (28)No space left on device: AH00023: Couldn't create the mpm-accept mutex

It was cleaned up using ipcs and ipcrm in a loop.

for line in `ipcs -s | grep apache | cut -d " " -f 2`; do ipcrm -s $line; done

Metadata Update from @rcritten:
- Issue assigned to akrivoka
- Issue set to the milestone: FreeIPA 3.3.x - 2013/10 (bug fixing)

5 years ago

Login to comment on this ticket.