Solaris

hi all,

On reading the docs, on fsck pass in /etc/vfstab ->

q1) how do we specify the to run the fsck in preen mode ( -o p ) inside /etc/vfstab ?

q2) if this parameter is use to determine whether fsck on the mount point / raw device is needed during boot time, how does the system bootup so fast ? i remember running fsck as a long process, but i didn't see any fsck at all when i rebooted my system .. ( i have a few mountpoint with fsck pass = 1)
(is the fsck running in background ?)

q3) lastly, if i set 1 ,2 ,3 ,4 individually for my different mountpoints in "fsck pass" in /etc/vfstab, they still get check sequentially right ? (the number does not implies the sequence of the check)

Hope to hear from you gurus..

Regards,
Noob

I think the fsck_pass field is historical.
Solaris 9 and later do not use it, instead check all ufs file systems with "-o p".
The / /usr /usr/platform /var /var/adm are checked in sequence (in run level S, and of course only if present in vfstab), the other local filesystems are checked in parallel (in run level 2).
(Solaris processes run level S and 2 before it goes to 3.)
--
On journaled ("logging") ufs file systems, fsck checks the integrity of the intent log; if okay it skips a full check.
Since Solaris 9 "logging" is default, i.e. does not need to be in vfstab field 7; check with

Code:

mount -v | grep -w ufs

Hi MadeInGermany,

Thanks for your reply.

Q1) Does it mean that it is due to logging which is the reason why fsck can be completed in short time ?

Q2) If I have a filesystem/device with nologging, does that means fsck will take a longer time to complete and the system can only go to runlevel3 when the fsck is completed ?

Q3) when fsck is running on a filesystem/device, are we still able to access the filesystem ?

if the answer is no -> it means that fsck on / have to be completed before the system can boot into S user mode ?

Regards,
Noob

Q1) Yes, the main purpose of "logging" is a fast fsck.
Q2) Yes, much longer. The more files are on the file system, the longer it takes. Yes, run level 3 is reached when all fsck are done.
Q3) No, not accessable. fsck is to be run before the file system is mounted.
When the system is booted, / is mounted read-only, then fsck is run on it, then it is remounted read-write. (When fsck fails it goes to an emergency maintenance shell; / is still read-only then.)

If your goal is a faster system boot, simply remove the nologging option (field 7) in /etc/vfstab for the respective ufs file system (field 4)
If you have Solaris 8 you must put logging option (i.e. replace the - placeholder with logging).
Next boot (or next mount) will automatically convert the filesystem to logging (that can a few seconds, and it will reduce the free space by a few percent).

Hi MadeInGermany,

Thanks for your reply.
In the above, you mentioned that the filesystem is not accessible/mounted until fsck is completed.
But following you mentioned that / is mounted as read-only and fsck is being run on it.

So if a filesystem is mounted as read only, we can access the filesystem while fsck is running on it ?

What will actually happen if userA run a fsck on a mountpoint (e.g /mnt), and userB try to access the mountpoint (cd /mnt; ls /mnt ) ?

Regards,
Noob

Yes / is an exception, it needs to be there in order to have any files.
If a file system is read-only, you can run fsck on it. But as soon fsck modifies something, the system is likely to crash (because cashed things do not match real things), so it should be umounted/remounted ASAP.
If a file system is read-write, you can technically run fsck on it. But if either a user or fsck does modifications during the fsck run, the whole stuff goes out of sync, and the system is likely to crash with a permanently inconsistent file system. Once inconsistent, any further write will grow the inconsistencies. A long grown inconsistency can lead to a point that finally an fsck during a system boot will move a major part of the files to lost+found, with lost attributes, file names, directory names, directory hierarchy.