[EMAIL PROTECTED] writes:
> To: [EMAIL PROTECTED]
> Subject: DB Backups (was Re: AFS Disaster Recovery ...)
> From: William Setzer <[EMAIL PROTECTED]>
> Reply-To: [EMAIL PROTECTED]
> Date: 30 Dec 1999 14:18:47 -0500
> In-Reply-To: [EMAIL PROTECTED]'s message of Mon, 27 Dec 1999 14:12:36 -0500
>(EST)
>
> [EMAIL PROTECTED] writes:
> :
> : Saving the UBIK databases requires an administrator to shut down the
> : database servers on a single database machine (preferably one that
> : is not the sync site), copy the raw database files on that machine
> : "/usr/afs/db/*.DB0", start the database servers, and then store/save
> : (tar) the copied databases along with the "/usr/afs/etc/" directory
> : to tape.
>
> I've always wondered why this isn't easier. Has anyone tried to
> yank out the ubik urecovery stuff into its own little binary that
> lets a non-db suck down archive copies?
>
>
> William
>
Sure.
/afs/umich.edu/group/itd/build/mdw/uniqname2.0/src/unfetch.c
It will need some work if you want to use it for something else -- the
source has port 7010 wired in, because we use it to back up uniqname.
Note that while it's running, it will block any write
transactions on the database until it's done. I seem
to recall this can also cause reads to block behind
the write on the sync site. This is probably not a concern
unless you have a large or very busy site.
-Marcus Watts
UM ITD PD&D Umich Systems Group
