Difference between revisions of "Intervention report 2017-01-12"

From Pumping Station One
Jump to navigation Jump to search
Line 38: Line 38:
 
= How to avoid the problem in the future =
 
= How to avoid the problem in the future =
  
- Well I had to learn how things worked on the fly. We plan on updating the
+
- Well I had to learn how things worked on the fly. We plan on updating the infrastructure to more maintenable systems (arch was a big problem here as it was not updated for a long time).
  infrastructure to more maintenable systems (arch was a big problem here as it
 
  was not updated for a long time).
 
  
 
- Have emergency clone VMs to restore quickly (if one has to leave before finishing upgrade/maintenance etc)?
 
- Have emergency clone VMs to restore quickly (if one has to leave before finishing upgrade/maintenance etc)?

Revision as of 17:18, 12 January 2017

Description

Login was broken on all machines of the space and ps1auth was complaining a lot about other stuff. Took over someone that was trying to correct it but had to leave. At 2017-01-12T20:00:00

Intervention by

- bjonnh

Problem

- Seen some libraries problems on the ps1auth VM. Tried to update them. Then started the clusterfuck.

Resolution

Ended at: 2017-01-12T22:00:00

- Did a full update of arch on ps1auth.

- Only some stuff got through

- Had to disable GPG signing of packages for that using sed to modify pacman.conf, because nothing else was working

- Did the update

- Rebooted bob, resolved the initial problem of machines not logging, but I had to do that two times

- Had to reconfigure network because interface name changed

- As python was updated had to recreate the virtualenv of ps1auth

- Had to update some packages (beautifoulsoup4, kombu add django-ldap something) because they failed with the new python

- Rebooted the machine to be sure everything was ok

- Got home.

How to avoid the problem in the future

- Well I had to learn how things worked on the fly. We plan on updating the infrastructure to more maintenable systems (arch was a big problem here as it was not updated for a long time).

- Have emergency clone VMs to restore quickly (if one has to leave before finishing upgrade/maintenance etc)?