how long is reiserfsck expected to run?

how long is reiserfsck expected to run?

[jp]

Hi all,

i had a bad crash yesterday morning. I managed to get the raid array together 
and ran reiserfsck --fix-fixable on it.

By now it is running for 30h (500gb mail spool). How much more can i expect it 
to run? How can i get some indication whether it's 10% done, 50% done or 90% 
done?

--

Jure Pecar

Re: how long is reiserfsck expected to run?

[Oleg Drokin]

Hello!

On Sat, May 24, 2003 at 11:31:40AM +0200, jp wrote:

> By now it is running for 30h (500gb mail spool). How much more can i expect it 
> to run? How can i get some indication whether it's 10% done, 50% done or 90% 
> done?

It can take pretty long time.
(it prints percentage for each pass.)
But on the other hand we have some useful optimisations that should make this process
much more fast in cases like you describe.
This code is being tested by our testers, but it seems to be pretty
stable and numerous people tried it without any bad effects.
I can send you the latest version if you want to try it.

Bye,
    Oleg

Re: how long is reiserfsck expected to run?

[Jure Pecar]

On Sat, 24 May 2003 13:37:10 +0400
Oleg Drokin <green@namesys.com> wrote:


> It can take pretty long time.

It just finished, after about 40h.

2859 found corruptions can be fixed only during --rebuild-tree.

yay. And manual page strongly suggest backing up the partition before
--rebuild-tree. But i neither have 500gb handy nor time to do that ...

Btw, how long is --rebuild-tree going to take? Even longer i guess? :)



Anyway, here's what happened:

I have a 28 fs disks in a raid10, one stripe across 14 mirrors. Disks get
randomly kicked out of mirrors every now and then due to a driver/firmware
issue i'm currently hunting down with qlogic. So on wednesday evening sdh
got kicked out of md8, on early friday morning sdw, the other disk in this
mirror, also started to act. I did a failover to the other box in order to
sort things out, but, the horror, sw raid deceided to take sdh as a more
recent disk than sdw, thus overwriting fresh data on sdw with ~two days old
data from sdh.

I was able to get the stripe together normally, reiserfs mounted without
complains, but it was obvios immediately that something is wrong ... there
were files missing that should be there, touching them returned -EPERM.
Cyrus was complaining about its databases and didnt start ... Then i
unmounted the partition, tried reiserfsck --check, gave up after one hour
and went on with --fix-fixable ... 

Has anyone been through something like this before? Is this fixable
afterall?

Well if it is i think i owe you (the reiserfs team) something ... what do
you prefer? :)


Btw this is on rh AS with their 2.4.9-e16 kernel, based on 2.4.9-ac10. I'd
say that reiserfs code in this kernel is exactly the same as in the Alan's
ac10. I'm using reiserfsprogs 3.6.6.

-- 

Jure Pecar

Re: how long is reiserfsck expected to run?

[Oleg Drokin]

Hello!

On Sat, May 24, 2003 at 10:23:24PM +0200, Jure Pecar wrote:
> > It can take pretty long time.
> It just finished, after about 40h.
> 2859 found corruptions can be fixed only during --rebuild-tree.
> yay. And manual page strongly suggest backing up the partition before
> --rebuild-tree. But i neither have 500gb handy nor time to do that ...

Well, reiserfsck --rebuild-tree is much more robust these days.
The only thing you should not have on tihs fs is images of other reiserfs
filesystems in plain form (not compressed or encrypted).

> Btw, how long is --rebuild-tree going to take? Even longer i guess? :)

Generally speaking yes. But with the version I sent you it should be
faster. It would be interesting to know how long rebuild-tree run of
new reiserfsprogs will take

[parts of old data sticked into fs horrors]

> Has anyone been through something like this before? Is this fixable
> afterall?

It should be somewhat fixable, I think.
Some of the files (esp. aleady deleted ones) may easily end-up in
lost_found, though.

> Well if it is i think i owe you (the reiserfs team) something ... what do
> you prefer? :)

we have this http://namesys.com/support.html page that describes ways to
give us money if you feel the need to ;)

> Btw this is on rh AS with their 2.4.9-e16 kernel, based on 2.4.9-ac10. I'd
> say that reiserfs code in this kernel is exactly the same as in the Alan's
> ac10. I'm using reiserfsprogs 3.6.6.

This kernel is buggy and generally is not recommended for reiserfs operations.
I wonder if you can easily stick in at least some 2.4.19 flavour?

Also please use reiserfsprogs version I sent you.

Bye,
    Oleg