Description of problem: Using e2label to label an ext3fs on an iscsi LUN and then trying to mount by that lable fails. Version-Release number of selected component (if applicable): (I have no idea if this is actually an e2fs problem or an isci-initiator-utils issue, or something else entirely.) e2fsprogs-1.35-12.3.EL4 iscsi-initiator-utils-4.0.3.0-3 (System as a whole just updated to RHEL4u3 a few hours ago.) How reproducible: Very Steps to Reproduce: 1. #e2label /dev/sdb iscsi-jfk0 2.# e2label /dev/sdb iscsi-jfk0 3.# mount -L iscsi-jfk0 Actual results: mount: no such partition found (or, from an fstab entry of LABEL=iscsi-jfk0 /netapp/iscsi/jfk0 ext3 defaults,_netdev 1 2 'mount -a' gives [root@jfk /]# mount -a mount: special device LABEL=iscsi-jfk0 does not exist Expected results: A mounted filesystem Additional info:
Is '/sbin/findfs LABEL=iscsi-jfk0' working for you?
Seems so, yes: # /sbin/findfs LABEL=iscsi-jfk0 /dev/sdb (I'll have to switch testing to a different box for anything involving mounts/unmounts, as this is a production box that's hard to limit access to... So let's try the same things on my test box right now: [root@comfort ~]# e2label /dev/sdc1 test1 [root@comfort ~]# e2label /dev/sdc1 test1 [root@comfort ~]# /sbin/findfs LABEL=test1 /dev/sdc1 [root@comfort ~]# mount -L test1 /netapp/iscsi/jfktest/1 [root@comfort ~]# mount . . . /dev/sdc1 on /netapp/iscsi/jfktest/1 type ext3 (rw) { Oh My! I have a theory here... I'll expand at the end of this comment } [root@comfort ~]# umount /netapp/iscsi/jfktest/1 { I go and edit fstab ... } [root@comfort ~]# grep jfktest /etc/fstab LABEL=test1 /netapp/iscsi/jfktest/1 ext3 defaults,_netdev 1 2 [root@comfort ~]# mount -av . . . /dev/sdc1 on /netapp/iscsi/jfktest/1 type ext3 (rw,_netdev) So, my theory is that the important difference is: /dev/sdc1 (a labeled sub-partition of the iSCSI LUN/'device') works. /dev/sdb (A filesystem created using the whole LUN/'device' - which mkfs confirmed I meant to do, and I did.) does NOT work. Does that make any sense? If it's the case, I still think it's a bug that needs to be fixed - or at the very least documented, but at least it's something I can work around (on future systems) until it's fixed.
Yeah, I think the device vs partition thing is the right direction: [root@comfort jfk0]# e2label /dev/sdb test_label [root@comfort jfk0]# cd .. [root@comfort iscsi]# umount /netapp/iscsi/jfk0/ [root@comfort iscsi]# mount -L test_label !$ mount -L test_label /netapp/iscsi/jfk0/ mount: no such partition found [root@comfort iscsi]# findfs LABEL=test_label /dev/sdb
This is no e2fsprogs problem, then. Assigning to util-linux which provides /bin/mount.
This is actually a reasonably high-priority issue, as device id is not guaranteed for iSCSI LUNs and so, should my system reboot, I have no guarantee it will come back up correctly without manual intervention and perhaps significant downtime.
Hmm.. the mount command in RHEL4 checks partitions only (/dev/name[0-9]*). It seems like a bug. Thanks for your report.
Oh, I know it's a bug. :) I was just hoping to get the priority raised - and thus the time-to-resolution hopefully lowered. Of course, I can do that myself, and if someone lowers it again, I'll complain. It's all good.
This request was evaluated by Red Hat Product Management for inclusion in a Red Hat Enterprise Linux maintenance release. Product Management has requested further review of this request by Red Hat Engineering, for potential inclusion in a Red Hat Enterprise Linux Update release for currently deployed products. This request is not yet committed for inclusion in an Update release.
Regarding the suggestions in comment #3 that this should be documented - there is already a reference to this type of problem in the RHEL4 release notes: "Red Hat requires that all block devices be initialized with a valid partition table, even if there is a single partition encompassing the entire device. This requirement exists to prevent potential problems caused by erroneous or unintended partition tables on the device." http://www.redhat.com/docs/manuals/enterprise/RHEL-4-Manual/release-notes/ The easy work around for this until mount supports labels for whole disks is to use just such a partition table - I use these on iSCSI test boxes here for GFS & cluster testing and it works fine.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2007-0235.html