[lustre-discuss] lustre-discuss Digest, Vol 122, Issue 9

Chris Hunter chuntera at gmail.com
Fri May 6 08:25:53 PDT 2016


Yes e2fsck recovery is not performance optimized :(

I found enabling drive readahead was minor help for e2fsck recovery.

Yes ll_recover_lost_found_objs is only for OST volumes, sorry.

Recovery mdt files is time consuming. If you run lustre 2.7 or newer, lfsck
can automate some recovery.

Most mdt files the extended attributes and not block data is important but
e2fsck does not recognize this relationship.

Keep mdt e2fsck logs, it can help recover mdt files back to original
location.

chris hunter
chuntera at gmail.com
Thank you Chris.

I'm running e2fsk, the last e2fsprogs version, in the combined mds / mdt
and I think ll_recover is for OSTs. Am I wrong?

Currently my main problem is the downtime. After three days e2fsck is still
running and I see old messages in the list about e2fsck running for
weeks...I can't wait weeks to finish the e2fsck

More information about our lustre system: combined mds / mdt has 189 GB and
8.9 GB used. It was formatted with the default options.

Regards.

=============================================
Fernando Pérez
Institut de Ciències del Mar (CMIMA-CSIC)
Departament Oceanografía Física i Tecnològica
Passeig Marítim de la Barceloneta,37-49
08003 Barcelona
Phone:  (+34) 93 230 96 35 <(+34)%2093%20230%2096%2035>
=============================================

El 6 may 2016, a las 15:13, Chris Hunter <chuntera at gmail.com> escribió:

On 05/05/2016 04:05 PM, lustre-discuss-request at lists.lustre.org wrote:
Hi Fernando,

When e2fsck has finished, you should run "ll_recover_lost_found_objs"
command. This tool can recover files from OST lost+found directory using
file extended attributes.

https://build.hpdd.intel.com/job/lustre-manual/lastSuccessfulBuild/artifact/lustre_manual.xhtml#dbdoclet.50438219_44971


FYI, e2fsck is part of e2fsprogs rpm package. You can update e2fsprogs
independently from lustre version.

-- 
regards,
chris hunter
chuntera at gmail.com

Date: Thu, 5 May 2016 15:37:13 +0200

From: Fernando Perez <fperez at icm.csic.es>

To: "lustre-discuss at lists.lustre.org"

   <lustre-discuss at lists.lustre.org>

Subject: [lustre-discuss] Problems running e2fsck mdt/mds: long time

   running it and a lot of errors

Message-ID: <572B4C89.9060108 at icm.csic.es>

Content-Type: text/plain; charset=utf-8; format=flowed


Hi,


We have a lustre 2.4.1 filesystem with 250 TB, 158 TB used, and a

combined MDT/MDS with ldiskfs and the latest e2fsprogs


Due to problems mounting an OST and other hardware problems in the past

with other OSTs, we have started to run a e2fsck -fy in the combined MDT

/ MDS two days ago because after run e2fsck -fn I saw a lot of errors.

Previously I made a file level mdt / mds backup.


It seems  that the e2fsck is repairing all the inodes. Currently is

running pass 4 and returns for each inode this kind of messages:


Unattached inode 26977505

Connect to /lost+found? yes


Inode 26977505 ref count is 2, should be 1.  Fix? yes


Is it normal? What time can I expect to finish the e2fsck?


Regards.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20160506/f59b6cf3/attachment-0001.htm>


More information about the lustre-discuss mailing list