[Lustre-discuss] Excluding dead OST via exclude=
Jonathan Buch
jonathan.buch at hs-karlsruhe.de
Mon Oct 18 03:42:36 PDT 2010
Hello group,
I'm trying to exclude a dead OST but I'm having trouble. When I use
`df`, it hangs until I use `lctl --device "" deactivate`.
Should the exclude= mount option prevent df from locking up?
Environment:
* Lustre 1.8.3 RHEL5 on server-side
* Lustre 1.8.2 patchless client-side
From reading mailing-list entries and the Lustre 1.8 documentation, the
deactivating has to be done on the server as well as the clients. The
server-side deactivation seems only to forbid creation of new objects
on the OST.
The OST shows up as an inactive device with `lfs` on the client after
mounting (regardless of mount options).
mount -t lustre -o localflock,exclude=eu01-OST0003
192.168.21.173 at tcp1:/eu01 /tmp/mnt
When mounting with the exclude= option, it should disable the OST
completely on the client. My problem now is, that the option does not
seem to do what I expect.
Now, should the exclude= option prevent df from locking up?
Jonathan Buch
# lfs df -h
UUID bytes Used Available Use% Mounted on
eu01-MDT0000_UUID 24.4G 3.4G 19.7G 13% /tmp/mnt[MDT:0]
eu01-OST0000_UUID 3.5T 3.1T 224.0G 88% /tmp/mnt[OST:0]
eu01-OST0001_UUID 3.7T 3.2T 313.5G 86% /tmp/mnt[OST:1]
eu01-OST0002_UUID 3.7T 3.3T 146.0G 91% /tmp/mnt[OST:2]
eu01-OST0003_UUID : inactive device
eu01-OST0004_UUID 889.2G 488.6G 355.4G 54% /tmp/mnt[OST:4]
eu01-OST0005_UUID 6.3T 2.3T 3.7T 36% /tmp/mnt[OST:5]
eu01-OST0006_UUID 6.3T 2.5T 3.4T 40% /tmp/mnt[OST:6]
filesystem summary: 24.2T 14.9T 8.1T 61% /tmp/mnt
(I'm sorry for reposting this, but the same mail earlier this year
did not attract enough attention for anyone to comment)
--
---------------------------------------------------------------
Jonathan Buch, B.Sc.
Karlsruhe University of Applied Sciences
Institute of Materials and Processes (IMP)
Moltkestrasse 30
D-76133 Karlsruhe
Germany
E-mail: jonathan.buch at hs-karlsruhe.de
Phone: +49 721 925 1415
Fax: +49 721 925 1503
http://www.iaf.hs-karlsruhe.de/ice
----------------------------------------------------------------
More information about the lustre-discuss
mailing list