<html>
<head>
<meta content="text/html; charset=windows-1252"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">Il 22/02/2017 18:31, Deepak Naidu ha
scritto:<br>
</div>
<blockquote
cite="mid:FF114F53-CC11-4860-9521-95E9FB5DFDD5@nvidia.com"
type="cite">
<div>I have a setup where storage nodes use network-1 & client
nodes use network-2. </div>
<div id="AppleMailSignature"><br>
</div>
<div id="AppleMailSignature">In both server/client, I use
/etc/hosts entry to define storage node name, example node1,
node2, node3 etc... </div>
<div id="AppleMailSignature"><br>
</div>
<div id="AppleMailSignature">When client uses node1 hostname it
resolves to <span style="background-color: rgba(255, 255, 255,
0);">network-2 & when storage uses node1 it resolves
to network-1.</span></div>
<div id="AppleMailSignature"><br>
</div>
<div id="AppleMailSignature">I don't think GlusterFS has as many
scrubbing jobs like other SDS where they need cluster
interconnect, it's mostly client server replication/translator
model. Other use case where "different" network is helpful , if
you have remote replication(geo) going across the pipe.<br>
</div>
</blockquote>
<br>
I thought about this, but if servers are also clients, then this
would not work.<br>
Well, I suppose this is not possible or not required.<br>
<br>
Is it really the client which replicates the data and distributes it
to the different nodes?<br>
<br>
Alessandro<br>
<br>
<div class="moz-signature"><br>
</div>
</body>
</html>