[Cinder.glusterfs.ci] anyone using glusterfs as openstack block storage?

Zhu,Chao zhuchao at gmail.com
Sat Dec 20 08:08:34 UTC 2014


we are using glusterfs 3.4 as openstack block storage and are having two
severe limitations as below;
would like to get in touch with industry peers on this;

1. problem with thin provision and online rebalance issue;
   for the vm image(say 100gb), it is stored as a single file  in a single
brick(say 1TB), the brick); With thin provision,  the vm image file maybe
10g in the first week ,  and cinder may put 20 vm images on this
brick(still 10gb*30=300gb size); With vm growing after a while, each vm is
at 33gb and it is supposed to be able to grow th 100gb; But we run out of
disk space on this brick; -----so we have to rebalance; with rebalance, we
found some IO heavy VM got read-only while we do rebalance;

2. one node goes offline, full sync after it goes online;
  Say we have 3 node glsuterfs cluster, with 1TB storage each(one brick of
1 TB filesystem, each with 600gb storage used); with one node down for 10
min, and the files are consistently modified by VM(as vm always running),
when the node goes back online, all the vm images are different from other
2 copy vm files; so we have to re-sync the full 600gbfiles from other 2
nodes?  what is the exact logic behind it does the re-sync? 

Thanks for sharing your thoughts;
-- 1.
Regards
Zhu Chao
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/cinder.glusterfs.ci/attachments/20141220/fd6d192e/attachment.html>


More information about the Cinder.glusterfs.ci mailing list