Red Hat Bugzilla – Bug 594649
Read-only root FS with iSCSI target
Last modified: 2010-11-10 15:18:59 EST
Created attachment 415615 [details]
boot.log and messages
Description of problem:
lvm lv '/' belongs to vg, it is spread across local drive and iSCSI target.
'/' is read-only (mount says 'rw', however /proc/mounts claims correct 'ro').
Impossible to remount with 'rw'.
Version-Release number of selected component (if applicable):
Tested in kvm virtual machine with scsi-target-utils providing target
There are no errors in that log.
So, this is not a bug or it just doesn't manifest itself in the logs?
The latter; the logs do not show any data as to why it's happening.
This happens because of network init script. eth0 (device used for iscsi) is using dhcp, network script is starting it again, connection to iscsi target is lost, pv of lvm is lost, integrity of lv is lost, lv is remounted as 'ro'.
Bringing up interface eth0: EXT4-fs error (device dm-0): ext4_journal_start_sb: Detected aborted journal
EXT4-fs (dm-0): Remounting filesystem read-only
Determining IP information for eth0... done.
[ OK ]
Is the lease file properly ending up in /dev/.dhclient-<foo>.leases or /dev/.initramfs/net.<foo>.lease ?
(In reply to comment #5)
> Is the lease file properly ending up in /dev/.dhclient-<foo>.leases or
> /dev/.initramfs/net.<foo>.lease ?
How can I check that? Executing this on running system:
# find /dev | grep lease
didn't find anything
When running boot with rdbreak rdshell options, I can find these files in ramfs:
but this is not probably much useful information.
Basically issue occurs if ifcfg-eth0 contains:
Anyway, it isn't so simple, it is little bit weird, once issue occurs 'no' for these options isn't enough, some (nearly random) magic has to be done with disabling/enabling network, iscsi and iscsid services. Hard to say if it is caused by system or testing evironment.
CC'ing Harald about lease location... they should be ending up in /dev/.initramfs/ for initscripts to find them properly, not /tmp. (Which may still be happening after the rdshell invocation).
(In reply to comment #7)
> CC'ing Harald about lease location... they should be ending up in
> /dev/.initramfs/ for initscripts to find them properly, not /tmp. (Which may
> still be happening after the rdshell invocation).
first they are in /tmp and get copied over to /dev/.initramfs
After double checking the scripts, I figured, that the "copy-script" is not installed due to a bug.
Working as expected in RHEL6.0-20100701.0
Red Hat Enterprise Linux 6.0 is now available and should resolve
the problem described in this bug report. This report is therefore being closed
with a resolution of CURRENTRELEASE. You may reopen this bug report if the
solution does not work for you.