[Linux-cluster] "fsck" on large file systems at boot time

[Bob Peterson]

----- Original Message -----
| Good morning;
| We have a critical Oracle application running on a two node Red Hat
| clustered environment. (RHEL5u5)
| 
| 
| Red Hat clustering has worked extremely well for us; we have achieved
| better performance and improved reliability at a substantial reduced
| cost.
| 
| 
| The issue is that when the system reboots, it takes over an hour to
| perform
| an “fsck” on a 1.8T gfs file-system. We cannot afford this down-time.
| 
| 
| Is the “fsck” at boot time required on a gfs file-system?

Hi,

No, fsck is _not_ required for gfs at boot time.  Offhand, I'd have to
say you do _not_ want to run gfs_fsck at boot time because then it does
not know whether the volume is currently mounted on other nodes.  If it
is mounted elsewhere, the fsck can cause damage.  Of course, there are
special caveats and circumstances, like open-shared root, single-node
gfs, and such.

When in doubt, call the Red Hat support people because they
usually know best.

Regards,

Bob Peterson
Red Hat File Systems