<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<div class="moz-cite-prefix">On 04/09/2014 09:24, M S Vishwanath
Bhat wrote:<br>
</div>
<blockquote cite="mid:540821DA.805@redhat.com" type="cite">
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
<div class="moz-cite-prefix">On 04/09/14 00:33, Vijaykumar Koppad
wrote:<br>
</div>
<blockquote
cite="mid:CAC1pKAVFoURsKuuj5a2C1+crSWG3jTWvECw2AvkXHiU9DipR9w@mail.gmail.com"
type="cite">
<div dir="ltr">
<div class="gmail_extra">
<div class="gmail_quote">On Wed, Sep 3, 2014 at 8:20 PM, M S
Vishwanath Bhat <span dir="ltr"><<a
moz-do-not-send="true" href="mailto:vbhat@redhat.com"
target="_blank">vbhat@redhat.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="">On 01/09/14 23:09, Paul Mc Auley wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex"> <br>
Is geo-replication from a replica 3 volume to a
replica 2 volume possible?<br>
</blockquote>
</div>
Yes. geo-replication just needs two gluster volumes
(master -> slave). It doesn't matter what
configuration master and slave has. But slave should be
big enough to have all the data in master.
<div class=""><br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
Should I stop geo-replication before adding
additional bricks? (I assume yes)<br>
</blockquote>
</div>
There is no need to stop geo-rep while adding more
bricks to the volume.
<div class=""><br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
Should I stop the volume(s) before adding additional
bricks? (Doesn't _seem_ to be the case)<br>
</blockquote>
</div>
No.
<div class=""><br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
Should I rebalance the volume(s) after adding the
bricks?<br>
</blockquote>
</div>
Yes. After add-brick, rebalance should be run.
<div class=""><br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
Should I need to recreate the geo-replication to
push-pem subsequently, or can I do that out-of-band?<br>
...and if so should I have to add the passwordless
SSH key back in? (As opposed to the restricted
secret.pem)<br>
For that matter in the inital setup is it an
expected failure mode that the initial
geo-replication create will fail if the slave host's
SSH key isn't known?<br>
</blockquote>
</div>
After the add-brick, the newly added node will not have
any pem files. So you need to do "geo-rep create
push-pem force". This will actually push the pem files
to the newly added node as well. And then you need to do
"geo-rep start force" to start the gsync processes in
newly added node.<br>
<br>
So the sequence of steps for you will be,<br>
<br>
1. Add new nodes to both master and slave using gluster
add-brick command.<br>
</blockquote>
<div>After this, we need to run "gluster system:: execute
gsec_create " on master node and then proceed with step
2.</div>
</div>
</div>
</div>
</blockquote>
Yeah. Missed it... Sorry :)<br>
</blockquote>
Ah, I suspect that's the step I was missing.<br>
<blockquote cite="mid:540821DA.805@redhat.com" type="cite"> The pem
files needs to be generated for the newly added nodes before
pushing them to slave. Above command does that.<br>
<blockquote
cite="mid:CAC1pKAVFoURsKuuj5a2C1+crSWG3jTWvECw2AvkXHiU9DipR9w@mail.gmail.com"
type="cite">
<div dir="ltr">
<div class="gmail_extra">
<div class="gmail_quote">
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex"> 2.
Run geo-rep create push-pem force and start force.<br>
3. Run rebalance.<br>
<br>
Hope this works and hope it helps :)<br>
</blockquote>
</div>
</div>
</div>
</blockquote>
</blockquote>
Thanks for that folks, additionally I had assumed I'd need to stop
the replication before rebalancing or redoing the replication. It
appears to be a bit more stable if I proceed as described.<br>
Also my read is that while replication between the master and slave
is between individual nodes which happen to host single bricks, that
master doesn't deal with the slave's bricks.<br>
Paul<br>
</body>
</html>