| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Cott Lang <cott(at)internetstaff(dot)com> |
| Cc: | pgsql-general(at)postgresql(dot)org |
| Subject: | Re: tablespaces a priority for 7.5? |
| Date: | 2004-01-22 16:04:30 |
| Message-ID: | 12421.1074787470@sss.pgh.pa.us |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-general |
Cott Lang <cott(at)internetstaff(dot)com> writes:
> However, unless I am mistaken, a baseline backup would need to be taken
> cold because you cannot take a consistent online backup of the data
> files without using file system snapshots or split mirrors, and even
> that's questionable.
No, it wouldn't. All you need is to archive WAL beginning from the last
checkpoint record before you begin to take the baseline backup. The
baseline backup would not be consistent in itself --- but replaying WAL
from the previous checkpoint to any time later than the end of the
backup would bring it to a consistent state. Missed updates in the
backup are essentially damage that would get repaired by replay.
There are a few trivial things that still have to be done before this
can be a reality (one I can think of is that WAL really needs to have
entries for file creation/deletion), but it's not that far off in terms
of the base mechanisms. Writing the management software is the main
task.
There is a lot more info available in the pghackers archives.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tom Lane | 2004-01-22 16:11:00 | Re: trust auth in 7.4 |
| Previous Message | David Garamond | 2004-01-22 16:02:29 | Re: embedded/"serverless" (Re: serverless postgresql) |