Thread: Non-simultaneous file system snapshots as backups
I know we allow people to use file system snapshots as backups, but what happens if they are using tablespaces and they can't do the snapshots simultaneously? If there is only one check point happening between the first and last snapshot, would the WAL logs clean up that inconsistency like they do for crashes? I assume the pg_xlog directory would have to be the last file system snapshotted. If so, is this something we should document? -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +
On Fri, Mar 11, 2011 at 17:46, Bruce Momjian <bruce@momjian.us> wrote: > I know we allow people to use file system snapshots as backups, but what > happens if they are using tablespaces and they can't do the snapshots > simultaneously? If there is only one check point happening between the > first and last snapshot, would the WAL logs clean up that inconsistency > like they do for crashes? I assume the pg_xlog directory would have to > be the last file system snapshotted. If so, is this something we should > document? If you can't take an atomic snapshot, you have to use pg_start_backup/pg_stop_backup. But as long as you do that, it works fine with any kind of snapshots. I don't think it's doable any other way. That said, there are systems that let you snapshot atomically across multiple tablespaces. But they tend to not be cheap. -- Magnus Hagander Me: http://www.hagander.net/ Work: http://www.redpill-linpro.com/
Magnus Hagander wrote: > On Fri, Mar 11, 2011 at 17:46, Bruce Momjian <bruce@momjian.us> wrote: > > I know we allow people to use file system snapshots as backups, but what > > happens if they are using tablespaces and they can't do the snapshots > > simultaneously? ?If there is only one check point happening between the > > first and last snapshot, would the WAL logs clean up that inconsistency > > like they do for crashes? ?I assume the pg_xlog directory would have to > > be the last file system snapshotted. ?If so, is this something we should > > document? > > If you can't take an atomic snapshot, you have to use > pg_start_backup/pg_stop_backup. But as long as you do that, it works > fine with any kind of snapshots. I don't think it's doable any other > way. > > That said, there are systems that let you snapshot atomically across > multiple tablespaces. But they tend to not be cheap. Agreed, thanks. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +