[p4] Backups interrupting Perforce?

Stephen Vance steve at vance.com
Tue Nov 7 11:10:09 PST 2006


Jeff --

It sounds like you are doing things correctly, although I'm a little 
surprised that it's affecting your developers after the checkpoint has 
been taken.

For the performance issue, you need to monitor the box to determine 
where the issue is. Perhaps the backup software uses lots of memory for 
caching or CPU for compression, or maybe it's about file system I/O or 
disk thrashing. Until you can pin it down, you're shooting in the dark.

That said, there are some optimizations to Perforce backups you can do 
regardless.

If you're willing to take the trade-offs of incremental backups, you can 
do those and only do full backups every week or whenever suits your pain 
threshold.

Many operating systems and filers have the ability to take file system 
snapshots. There are techniques for using these snapshots to increase 
your availability. While they're conceptually the same across 
implementations, the exact hardware makes it difficult to give 
specifics. Look at Richard Geigers script for NetApp filers in the 
Public Depot for the concept and one implementation. For one client, it 
took their offline time from almost 2 hours down to 20 seconds.

Steve

Jeff Desharnais wrote:
> We are having an issue here where our assets are growing larger and
> larger and thus our backup times (LTO tape) are increasing.
>
> Because of this, we are actually having the backup process creep into
> the mornings at times, where developers are starting their work day.
>
> When this happens, we are noticing hangups of Perforce as the backup
> finishes up.
>
> Obviously, this is something we want to prevent.
>
> We are currently backing up all of P4ROOT with the exception of the db.*
> files. We are also backing up the actual depot folders on the other
> drive.
>
> I suppose my question is twofold.
>
> 1) What backup procedures are recommended for backing up Perforce and
> the assets? Any suggestions or common practices you could share?
>
> 2) If what we are doing is the recommended way, are the hangups expected
> and are they avoidable?
>
> Thanks,
>
>
> ---------------------
> Jeff Desharnais
> IT Specialist
> Mad Doc Software
> 100 Burtt Rd., Suite 220
> Andover, MA  01810
> (978) 409-6272 Ext. 132
>
> _______________________________________________
> perforce-user mailing list  -  perforce-user at perforce.com
> http://maillist.perforce.com/mailman/listinfo/perforce-user
>
>   



More information about the perforce-user mailing list