[CentOS] unattended fsck on reboot

Timo Schoeler timo.schoeler at riscworks.net
Thu Feb 18 18:10:38 UTC 2010


------- Original message -------
> From:  <m.roth at 5-cent.us>
> To: centos at centos.org
> Sent: 18.2.'10,  18:55
>
> Timo wrote:
>> ------- Original message -------
>>> From:  <m.roth at 5-cent.us>
>>> Sent: 18.2.'10,  18:21
>>>> On Thu, Feb 18, 2010 at 10:12 AM, Henry Ritzlmayr
>>>> <fedora-list at rc0.at>wrote:
>>>>> nate wrote:
>>>>  >
>>>>> >    I don't bother changing the setting for local disks as it is
>>>>> >    usually pretty quick to scan them. You must have a pretty big
>>>>> >         and/or slow file system for fsck to take 2+ hours.
>>>>> >
>>>>> > This particular server has 2x 500GB HDD's with failry "full" XEN
>>>>> > VM's on it, each with it's own LVM volumes, so I guess it's a bit
>>>>> > more complex than a normal ext2 system :)
>>>
>>> Um, we've only got some cluster nodes with drives under 750G; most of
>>> ours have that, or 950G, and we're moving a lot to 1T, and some 1.5T.
>>> Then there's large raid arrays. They take a *while* to fsck. Unless I'm
>>> having serious disk problems, I've had to boot using fastboot as a
>>> kernel line parm for grub.
>>
>> What about using a 'decent' file system, such as XFS?
>
> Lessee, I'm *not* talking about my system at home.

Hm, I'm talking about both: My file server (about 10TiByte RAID) and...

> a) I'm talking about
> work;

..my systems at work.

> b) My manager, my co-worker, and myself support nearly 200 servers,
> including 5 clusters.

Roughly the same numbers here - and that's the CentOS boxen only, not to 
mention xBSD, Solaris, and AIX machines.

IMHO one of the main reasons for running (Open)Solaris for most people 
(read: those w/o having to run Solaris due to historical reasons) do so 
because of ZFS.

XFS has different feature sets, strenghts and weaknesses, but the advantage 
of both XFS and ZFS is their total lack of fsck hassles, compared to extN.

> Some people that we support "only" run jobs that go
> for 2-4 days, but there was the guy who was running a job that I had to
> wait until it was finished to reboot the NFS server with his home
> directory... and I waited ->two weeks<-.

So, where's exactly the connection to underlying file systems?
 
> Then there's the question of how we'd migrate.
>
> Sorry, time for a real world check.

One of my most favorite jokes is to begin IT related discussions with 'in 
an ideal world'... ;)
 
>          mark "that is, if my manager was willing in the
>                   first place"

Timo



More information about the CentOS mailing list