Ask Your Question
1

How to configure a shared file system in Openstack Compute? [closed]

asked 2015-12-16 06:44:12 -0500

zekken gravatar image

I am trying to implement NFS (shared /var/lib/nova/instances directory) in openstack compute so that when the host fails, we can still move the instance to other nodes. I am following the guide from: http://docs.openstack.org/admin-guide.... I tried to follow the steps in the mentioned link, and then landed up into issue that now I am unable to launch the instances. I rechecked everything and found that, there is a mismatch of UID/GID between my controller and the compute nodes (I am using two compute nodes). I tried manually to match them by following the guide mentioned in https://docs.hpcloud.com/commercial/G..., but still no help. I am trying to build a production level Openstack environment, but need to implement shared file system on the compute. Kindly provide a actually tested way to implement shared file system on the compute.

edit retag flag offensive reopen merge delete

Closed for the following reason the question is answered, right answer was accepted by Vinoth
close date 2016-01-09 03:07:11.804328

Comments

Hi, can you paste nova logs?

Eduardo Gonzalez gravatar imageEduardo Gonzalez ( 2015-12-16 06:57:35 -0500 )edit

There only error that I have found is : Caught SIGTERM, exiting

zekken gravatar imagezekken ( 2015-12-16 08:55:42 -0500 )edit

1 answer

Sort by ยป oldest newest most voted
1

answered 2015-12-17 00:44:14 -0500

Vinoth gravatar image

Hi,

I followed this link for my live migration setup. It worked fine.

Initially, I too faced the same UID/GID issue. In your case, Double check the UID / GID of nova and libvirt-qemu should be same in the both compute nodes. It may show the UID you are trying to assign is already in use. In that case, pick any one common unused UID from both the compute node and assigned it commonly in both the compute nodes.

Also, check the /var/lib/nova/ file and its owner & permission. Since you have changed the UIDs, there might be some chances for the nova UID is assigned to any other services that will get reflect in the file permission.

Thanks, Vinoth

edit flag offensive delete link more

Comments

I already followed this link, thank you anyways. I got the error only after trying to change the ids in both the controller and compute nodes. In my scenario, the compute nodes has the same uid/gid, but controller doesn't. The nova uid was the same as the neutron uid in compute, then this happened.

zekken gravatar imagezekken ( 2015-12-17 08:35:54 -0500 )edit

I have changed the uid/gid, but running this particular step : find / -uid 106 -exec chown nova {} \; and find / -uid 107 -exec chgrp nova {} \; is giving me error that "Operation is not permitted"

zekken gravatar imagezekken ( 2015-12-17 09:29:34 -0500 )edit

And now I am getting this error: TRACE nova.compute.manager [instance: 67c7f9be-17c0-4aca-9634-eb95fbe52c6a] error: Unable to acquire lock on /var/lib/nova/instances/locks/nova-f39e81e64e7df18402c91c41da2e0adcd27a0c67 due to [Errno 37] No locks available

zekken gravatar imagezekken ( 2015-12-17 11:20:35 -0500 )edit

Yes, That's why I suggested you to use common UID/GID for both the servers. Reboot the servers and check the UID/GID for both. Then you must have UID/GID as same in both the servers.

"Operation is not permitted" -- I too got this error but still live migration worked for me.

Vinoth gravatar imageVinoth ( 2015-12-18 01:18:41 -0500 )edit

find / -uid 106 -exec chown nova {} \;

The above command is for changing the owner to the relevant folders. So you could manually change the owner permission.

Vinoth gravatar imageVinoth ( 2015-12-18 01:20:14 -0500 )edit

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2015-12-16 06:44:12 -0500

Seen: 1,575 times

Last updated: Dec 17 '15