Red Hat Bugzilla – Bug 1287316
volumes not mounted fast enough resulting in system boot to emergency mode.
Last modified: 2017-01-30 06:08:53 EST
Can you boot the machine with debug on kernel cmdline, reproduce the issue and send us output of journalctl?
Logs with debug are attached so removing needinfo ..... Customer is active so we can collect or try out things.
To me this looks like a dm/lvm bug. The system was not able to assemble the raid.
But anyway, even if this is a systemd bug, we will need some help from storage guys, so reassigning to lvm.
Are those failing mount points on LVM and then on multipath stack?
(In reply to Peter Rajnoha from comment #10)
> Are those failing mount points on LVM and then on multipath stack?
Yes, In my case mount points are on LVM which are using multipath device.
I have latest sosreport from customer, I can upload here.
(In reply to ajit mote from comment #11)
> (In reply to Peter Rajnoha from comment #10)
> > Are those failing mount points on LVM and then on multipath stack?
> Yes, In my case mount points are on LVM which are using multipath device.
> I have latest sosreport from customer, I can upload here.
If the customer uses lvmetad (use_lvmetad=1 set in lvm.conf), if possible, it would be great if he could try suggestion from bug #1287106 comment #47.
If he doesn't use lvmetad (use_lvmetad=0), then bug #1287106 comment #42 and bug #1287106 comment #43.
I think bug #1287106 and this bug may be about the same.
There are device-mapper-multipath test packages that avoid reloading a multipath device until multipathd has received the uevent from creating it. You can download them here:
These packages, along with the change to pvscan that Peter posted in bug #1287106 comment #48, should hopefully fix this (assuming that this is the same issue as bug #1287106)
Is this bug report linked with this solution (I've access to this one) ?
I've tried this solution in our environment and it seems to be a good workaround for this problem. But we don't want to put in production (Oracle RAC on RH Linux, so definitively a high availability platform ) something which has not a definitive fix, is following this bug report the right thing to do ?
I did'nt have access to bug report #1287106...
(In reply to LENHOF from comment #46)
> I did'nt have access to bug report #1287106...
There's customer data scattered through that bug so I cannot easily open it up unfortunately.
We'll see if the information is somewhere else or if we need to copy it across here.
I've opened a new BZ with summary of the problem which is public:
So the original case is closely related to broken installation procedure provided by 'openstack undercloud install'.
It should not mess with 'auto_activation_volume_list'
As for the comment 58 - from provided report - it looks unrelated to this bz, as logs shows evidence of 'qla2xxx [0000:44:00.0]-8038:0: Cable is unplugged...' - so it's more or less likely a hardware issue with multipath configuration.
If there is still any problem - please open new case.
Closing the BZ as the original case is closed.