RAID 5 recommendation

[bofh]

:-) i was kidding about the ram drive but true maybe for some things might be usefull

i know about the nfs downsides. we use a central ftp service for all webserver data goes by nfs to the right server (we wanted to decrease holes on the webserver so we dropped every direct access there)

besides the big overhead its working well.

but my question about the backup on a nfs drive is what will zimbra do.
i know the backup will be trash but will zimbra retry to save the file again?

i dont know how this incremental backup is really working - does it check existing data or simply go by internal log.

so the question is if one is trashed by nfs failure will it try to resend the backup on the next shedule? or means incremental data is lost ?

what if backup path stays at it is and a script do the upload / verify / then delete stuff? will zimbra ignore the fact that no other incremental backup is in the diretory and simply go ahead?

its not nessesary a money thing but since we already have a storage server (which is replicated to a different location) i wanna use him instead of something else - IF its reliable

[LMStone]

I knew you were kidding about the RAM drive; thought I'd tweak you just a bit though on that! :-)

If your NFS mount dies while zmbackup is running I don't know what would happen; you could just look at the script and see what's there for error handling.

But since zmbackup is called by a cron job, I would expect that if it fails, it just fails and you need to wait for the next running.

All the best,
Mark

[3RiversTechAdmin]

A good NFS implementation has no issues. I have seen great throughput and low latency with NFS over 10 GigE.

But this depends on a number of things:
1. A solid network layer
2. Sufficient NVRAM on the target
3. Sufficient spindles on the target
[4. If you target is a Linux server, have the NFS buffers increased, and exported files on a RAID10]

Some of these concerns are exactly the same as if the storage was DAS.
I run a lot of VMware VMs over NFS and get fantastic performance.

[bofh]

Originally Posted by 3RiversTechAdmin

A good NFS implementation has no issues.

when do so there can always be an issue. so thats why it would be good to know exactly how the backupscript react under certain circumstances

for example - host can be down - maintainance - usually in the night low traffic time exactly when the backup runs :-)
switch can be dead or down at the worst time - when backup runs
some things get reconfigured or moved to antoher host and the zimbra host isnt the first updated client - so good to know if that can be a desaster

and many other possibilitys special that one we dont even think about (that the one im really afraid of :-)

so its true in the normal world if you do things right nfs is like a rock - ife mine running over a year without connection loss or troubles or even restart or remount anything.

but you cant rely on that an an implementation. so you need a failsave plan/implemantation (or better the backupscript should have) with setups like that - that why im asking if anyone has anyidea in which way that backup is operating


if the script simply fail on connect loss and redoo everything next day its fine - you can write a simple watchdog get an notification that something is missing and make a manual backup if one day is a fail

but what if backupscript dont rewrite last day backup and go ahead only with new piece, so a piece of an incremental backup is missing, you may loose important parts of your backup so thats why i wanna know exactly how the backup script works and do things. to bad that i could not find any documenation about its behaivior in case x

[sadiq007]

I just installed zimbra 7.1.0 on CentOS 5.6 (64 bit) with OS on RAID1 and /opt on RAID5, so is it ok? or i need to move on RAID10 for /OPT ?

[phoenix]

Originally Posted by sadiq007

I just installed zimbra 7.1.0 on CentOS 5.6 (64 bit) with OS on RAID1 and /opt on RAID5, so is it ok? or i need to move on RAID10 for /OPT ?

All over the forums there are threads that detail what RAID levels are recommended (including this three year old thread you've posted in), you can also go to the wiki and search for the word 'raid'.