<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Nathanaël Blanchet wrote:
<blockquote cite="mid:538C5EE3.1000808@abes.fr" type="cite">
<meta http-equiv="content-type" content="text/html; ">
<font size="-1">Hi all,<br>
<br>
I created 3 clusters in unique datacenter. I installed </font><font
size="-1"><font size="-1">my hosts </font>to the first one after
activating the gluster service in the cluster tab. A expected otopi
installed then the vdsm-gluster rpm and everything were ok for this
first cluster. <br>
The other ones weren't initially gluster ready and I wanted to activate
the service after installing the host (so they were without
vdsm-gluster package). But checking that item didn't do anything for
these hosts, so that I had to manually install vdsm-gluster package.
Then, when creating a volume, the UI told me that hosts couldn't
communicate. So I had to manually probe them. This leads me to that
question : is it an expected behaviour? The whole gluster job
(installing appropriated package and probing peers in the cluster)
should be done thanks to the UI cluster tab "gluster service".<br>
A last question : why isn't it possible to probe host on different
clusters in the same datacenter, could be very useful.<br>
</font></blockquote>
After checking the checkbox for gluster services you'll need to
migrated all vms from a host and put it into maintenance then use the
re-install link and the needed packages will be downloaded and
installed and a peer probe will be done. After that reactivate the host
and proceed to the next one. One caveat found: chkconfig glusterd on
isn't executed. You'll need to do that by hand and also start the
service or restart the server while in maintenance mode.<br>
<br>
Joop<br>
<br>
</body>
</html>