snapshots on production best practice

[Paul Macdonald]

Hi,

We're recently come across an issue where server performance seems very 
tied to the number of ZFS snapshots on the box.

(deleting some snapshots has an instant effect on web server load times 
for example)

This is a system with 2 ZFS jails so likely exacerbated, however we tend 
to prune quite often, keeping for example (per zfs-jail) last 30 days, 
then weeklys perhaps going back 3-4 months, which doesn't seem outrageous.

Obviously this is all largely dependent on what actual disk IO is going 
on, but i'd be interested to hear what others recommend.

We also keep snapshots ( quite a lot) on the backup boxes which do no 
other work, without issue and thats maybe the answer right there,

the convenience of local snapshots  and the trivialness of creation has 
led us to let snapshots build up even on production boxes as we've seen 
no real general issues, but clearly there is work going on behind the 
scenes and performance must suffer.

Paul.

-- 

-------------------------
Paul Macdonald
IFDNRG Ltd
Web and video hosting
-------------------------
t: 0131 5548070
m: 07970339546
e: paul at ifdnrg.com
w: http://www.ifdnrg.com
-------------------------
IFDNRG
40 Maritime Street
Edinburgh
EH6 6SA
----------------------------------------------------
High Specification Dedicated Servers from £100.00pm
----------------------------------------------------