Incremental Backup Frequency

[cjstone]

I'm reconfiguring my backups a bit and have some questions...

First, some thinking out loud..
I've got /opt/zimbra/backup as a mountpoint for a different storage device than /opt/zimbra. This allows me to restore back to the last incremental should the /opt/zimbra device fail. Right now I do daily incrementals, so worst case is I'll loose 23+ hours of mail in event of disaster (not great).

I've always been backing up /opt/zimbra/backup to tape daily as well, so having this backup target gives me faster restores (don't have to go to tape now), but still the same potential amount of lost mail.

My first thought was to symlink /opt/zimbra/redolog to /opt/zimbra/backup/redolog. Zimbra would then write redologs to the other volume, giving me up-to-the message restore capability.

This sounds great, but now should the /opt/zimbra/backup volume fail, zimbra would stop writing redologs and that wouldn't be good, I'm sure.

My current thinking now, then, is to just run much more frequent inrementals, say every one or two hours. This keeps my redologs with /opt/zimbra until the backups, then gives only 1-2 hours of lost mail worst case, which is acceptable. (Our storage is hot-spared RAID 10 fibre SAN LUNs, so <whisper>a complete failure is not very likely. knock, knock.</whisper>, but I just need to be ready for anything).

My questions, then, are these:

1. Does anyone think symlinking /opt/zimbra/redolog to another volume is a good/bad idea?
2. What would be the drawbacks of running incrementals as often as hourly? As I understand it, with incs, the redologs just get moved over and restarted -- without going into maintenance mode, and it happens pretty quickly. (We have ~300 accounts, ~200GB mail).
3. Any other thoughts on how I can accomplish my goal of keeping my restores as up-to-date as possible should /opt/zimbra fail?

Thanks,

--Chris

[heinzg]

Hi is this regarding a NE or FOSS??

heinzg

[cjstone]

ZCS Version: Release 4.5.11_GA_1752.RHEL4_20080128123856 RHEL4 NETWORK edition, as shown in my profile

[heinzg]

ZCS Version: Release 4.5.11_GA_1752.RHEL4_20080128123856 RHEL4 NETWORK edition, as shown in my profile

WOOPS it is late here and I have my shades on

heinzg

[cjstone]

Any thoughts on this yet? Do most people just stick with the default incremental backup schedule in Network?

[JoeB]

cjstone - I'm in a similar situation as you. I want to reduce the data loss in the event of a failure.

I am considering running houlry incemental backups as well as a nightly full. What did you end up doing with your backups? Thanks!!!

[cjstone]

JoeB, I've been doing bi-hourly incrementals since soon after I posted, and it's been working fine--have noticed no stress on the servers.

I still do only weekly fulls, though, since /opt/zimbra/backups goes to tape each night, and nightly fulls would take too long (and use too much tape).

I have done test restores back from tape to confirm the backups, without problem.

--Chris

[JoeB]

Chris - Thanks for the info it will save me some time!