Talk:IT Development/Proposals/Infrastructure

From Wikimedia UK
< Talk:IT Development
Revision as of 17:03, 19 November 2012 by ErrantX (talk | contribs) (+)
Jump to navigation Jump to search

My remarks

I want to address 3 points:

  • We need IMO a permanent test VM to be able to test any change we want to implement before doing it on prod. (at least for the Web services like CiviCRM, Mediawikis, ...).
  • What about the backups of these VMs? I think we should have somewhere, something like rsnapshot running (incremental backup)? Or should we backup whole VMs? Do both?
  • What about a mirror for critical services? Do we need one? If not, what would be the recover strategy base on the backups in case of disaster? My opinion is that we should have a backup&mirror solution running somewhere else (other company, other place) in UK. Kelson (talk)
Certainly a backup server would be good; and really depends on the capacity required. The host usually has a feature for snapshots/VM backup so I don't think we need to worry about that. But if we have all the code/files (including configuration) in version control (and possible a provisioning system like Puppet) that could be backed up, along with SQL dumps etc. Also we discussed backups for a server in the office for files. So a standalone VM to dump all of that on sounds sensible. One of the problems with backing up to an external provider, however, is that data charges can begin to escalate... I'd suggest we have a main backup server within the same cloud account which takes regular syncs from the other servers - and we can set up a system to archive it to "cold storage" (something like Amazon glacier). In terms of hot mirrors with another host; I think that would be worth putting as a separate item after migration; once we have everything in source control etc. then it will be a bit easier to set that up. --ErrantX (talk) 14:56, 19 November 2012 (UTC)
Regular backups by the same company seems to me acceptable if the provider can garanty that the backups are not in the same place/city like the originals. OK to speak about mirroring later. Kelson (talk) 15:09, 19 November 2012 (UTC)

Past thinking

It might be worth having a look at 2012 Developer budget (in particular the 'Servers' section) for past thinking on infrastructure. Thanks. Mike Peel (talk) 12:37, 19 November 2012 (UTC)

Nearly there

I just have a couple more things to hear back on. Kelson, Mike, anyone else - comments on these suppliers?? Other supplier suggestions. --ErrantX (talk) 16:03, 19 November 2012 (UTC)