On 07/10/2017 10:15 AM, Stephen John Smoogen wrote: > On 10 July 2017 at 12:54, Dusty Mabe <[email protected]> wrote: >> >> >> On 07/10/2017 11:24 AM, Kevin Fenzi wrote: >>> On 07/07/2017 03:12 PM, Dusty Mabe wrote: >>>> >>>> As part of some ongoing work we are doing [1] to unify ostree repos >>>> I'd like to inquire about backup policy and data recovery for the storage >>>> that holds ostree repos within Fedora infra. The following directory is >>>> the subject: >>>> >>>> https://kojipkgs.fedoraproject.org/atomic/ >>>> >>>> Can someone tell me what we do to back those files up, how often we >>>> snapshot, >>>> how long the backups are kept and how we would go about restoring the >>>> content? >>> >>> This space is on our fedora_koji netapp volume. >>> >>> There's daily snapshots for 5 days currently. >> >> so backups are taken daily. cool. Is it possible to view what the snapshots >> are? >> i.e. readonly mountpoints on infra machines where content could be browsed. >> That >> would make requesting a restoration super easy since the requester could >> investigate >> by inspecting readonly directories before hand. This also may make it so >> that a >> restoration isn't needed at all for some cases where only a few files need >> to be >> restored. >> > > OK let us be careful about terminology here. Backups are usually > considered data which is taken of a system and held somewhere else to > be restored later. (EG a tape backup, a rdiff backup, etc). > > Netapp snapshots are not backups in that sense. What you get is the > ability to go into a .snapshot directory and look at what was present > at a certain point to certain limits (netapp snapshots are similar > (but not the same) to a hardlink list of data on disk at a certain > time. You can only change so much or keep so many before you run out > of snapshot space and those go away). > > Snapshots can be seen on certain machines by going into the .snapshot > directory.
I'll add that we disabled this on most machines. It caused various scripts to look at the snapshot data when it shouldn't have. We could re-enable it on some specific machines if needed, but right now it's off I think. > >>> >>> There's also a snapmirror from phx2 to offsite space in rdu2. >> >> What exactly does this mean? > > The Red Hat Netapps regularly copy the data from RDU2 to PHX2. As long > as we don't change too much over on PHX2, then the RDU2 side is within > sync with PHX2 within an hour. If we have a large delta or other > network traffic is full it can take many hours or certain snapshots > may get 'dropped' because they are no longer possible to reconstruct. Yeah, basically it means we have a copy (as up to date as sync time) in RDU2 of our PHX2 koji data. kevin
signature.asc
Description: OpenPGP digital signature
_______________________________________________ infrastructure mailing list -- [email protected] To unsubscribe send an email to [email protected]
