Quote:
Originally Posted by 6South
Really the important question is why are you having to run fsck on a 6TB volume in the first place? Having to run it at all indicates something seriously hit the shit, why is your server crashing so badly it's corrupting the file system?
6TB on a single machine is NOT a big deal though, I have machines with double that amount of data on them up for years with no issues. I've had to replace failed drives, rebuilt RAID arrays, etc but I've never had one down for hours running fsck.
I'd recommend adding a SAN or NAS to your setup given the amount of data though, at least you'd be on a platform that can resume service without a 6 hr file rebuild short of someone dropping a bomb on the server. A network storage system w/ replication would be your best bet in terms of ending downtime issues.
|
We suspect the problem is actually related to r1soft while performing hot copy/snapshots for backups on creating a larger than 2tb "back up chunk". If you want more information on how r1soft does its backups I can point you to their documentation. R1softs backup software only works with ext3 file systems, so using XFS file systems.. which is preferred on large file systems could not be backed up via r1soft due to the drivers that r1soft integrates into the kernel. Sadly Ext3/4 are both journalized file systems... and can be hell if you have to check the file system. Most file system errors that are easy to get resolved in these cases are due to inode issues/duplication which are easy to be fixed once the system boots into a single user mode and has access to the spindles to basically de-dupe the duplicate inodes.
A SAN.. stil doesn't help file systems, a SAN is just a storage level object that resides on a network, the access to the spindles is still block level. The file system still sits on top of this level.
A NAS.. same thing unless your talking a proprietary file system thats under the network storage protocol of NFS.. such as Netapp or, ZFS for Solaris or EMC's file system. Otherwise your just using a linux box with somesort of ext/xfs file system under a network protocol.