On 2014-04-25 03:35PM, Elliot Voris wrote: > I’m really interested to see what other admins are doing for > database/OS/code backups (especially curious about you big ol’ > libraries with huge clusters :-D).
At Sitka, we've got our primary database server, plus a standby server that is kept continuously up-to-date via streaming replication -- more or less as described here: http://www.postgresql.org/docs/current/static/warm-standby.html In our case, the standby is actually a hot standby server, which means that it can be used for some read-only queries as well as for backup purposes. We also have a process for daily backups, which are stored offsite in case the datacentre explodes. The backups are encrypted and there is a reasonable rotation scheme. As for our application servers, we generally don't back them up directly. Instead, we use git to manage code customizations, config files, etc. For example, when we want to make a change to Evergreen (e.g. apply a bugfix), we commit it to our main EG git repository, then deploy it consistently to our application servers. Likewise, if we need to modify our Apache config, we make the change to the relevant config files in git and push it out from there. I think of this as an ad hoc configuration management approach; we're planning to move towards more formal config management later this year, partly to capture aspects of our setup that the current approach leaves out (like OS-level stuff). -- Jeff Davis Lead Evergreen Specialist BC Libraries Cooperative
