This has to be asked in PVE mailing list, nothing to do with DRBD.<div><br></div><div>Check PVE documentation for how HA works (keyword: nofailback).</div><div><br></div><div>AFIAIK PVE's HA manager has no way to interact with the underlying storage. It will just attempt to start the resource wherever you have set it to do so.If the undetlying storage (DRBD in this case) is ready to do that, it will do it. I assume you are using DRBD8 in dual primary mode (with all risks that implies, if improperly configured).</div><div><br></div><div><br></div><div><br><br>On Thursday, August 9, 2018, Alexander Trilissky <<a href="mailto:alexander.trilissky@gmail.com">alexander.trilissky@gmail.com</a>> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Hello!<div><br></div><div>
<p class="MsoNormal" style="margin:0cm 0cm 0.0001pt;font-size:11pt;font-family:Calibri,sans-serif"><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(34,34,34);background:white">I have an asymmetric
ProxMox 5.2 cluster with two main nodes (node-A and node-B) and one small qorum
node (node-C).<span></span></span></p>
<p class="MsoNormal" style="margin:0cm 0cm 0.0001pt;font-size:11pt;font-family:Calibri,sans-serif"><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(34,34,34);background:white">My aim </span><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(31,73,125);background:white">is </span><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(34,34,34);background:white">to get a HA cluster
with DRBD-driven storage on main nodes. It should act </span><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(31,73,125);background:white">as follows</span><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(34,34,34);background:white">:<span></span></span></p>
<p style="margin:0cm 0cm 0.0001pt 36pt;font-size:11pt;font-family:Calibri,sans-serif"><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(34,34,34)"><span>1)<span style="font:7pt "Times New Roman""> </span></span></span><span lang="EN-US">In working state all nodes are up,
DRBD resources are UpToDate, guest VM <span style="color:rgb(31,73,125)">is </span>running
on one of main nodes.<span></span></span></p>
<p style="margin:0cm 0cm 0.0001pt 36pt;font-size:11pt;font-family:Calibri,sans-serif"><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(34,34,34)"><span>2)<span style="font:7pt "Times New Roman""> </span></span></span><span lang="EN-US">When Passive main node (B) fails,
guest VM works as usual on Active main node (A). When Passive main node
restores, DRBD resource synchronizes automatically.<span></span></span></p>
<p style="margin:0cm 0cm 0.0001pt 36pt;font-size:11pt;font-family:Calibri,sans-serif"><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(34,34,34)"><span>3)<span style="font:7pt "Times New Roman""> </span></span></span><span lang="EN-US">When Active main node (A) fails, PM
HA restart<span style="color:rgb(31,73,125)">s</span> guest VM on Passive Main node
(B).<span></span></span></p>
<p style="margin:0cm 0cm 0.0001pt 36pt;font-size:11pt;font-family:Calibri,sans-serif"><span lang="EN-US">And
now the interesting:<span></span></span></p>
<p style="margin:0cm 0cm 0.0001pt 36pt;font-size:11pt;font-family:Calibri,sans-serif"><span lang="EN-US" style="font-family:Arial,sans-serif;color:rgb(34,34,34)"><span>4)<span style="font:7pt "Times New Roman""> </span></span></span><span lang="EN-US">If <span style="color:rgb(31,73,125)">the </span>rest
main node (B) fails or <span style="color:rgb(31,73,125)">was</span> gracefully shut<span style="color:rgb(31,73,125)"> </span>down, and after that we start <span style="color:rgb(31,73,125)">the </span>first main node (A), PM HA starts guest VM on
this node with old data on disk. So we lost data changes on <span style="color:rgb(31,73,125)">the </span>main node B.<span></span></span></p>
<p style="margin:0cm 0cm 0.0001pt 36pt;font-size:11pt;font-family:Calibri,sans-serif"><span lang="EN-US">Is
there any way to allow PM HA <span style="color:rgb(31,73,125)">to </span>start guest
only on <span style="color:rgb(31,73,125)">the </span>node with last activated
resource in case of not fully mashed cluster, and to disable this if <span style="color:rgb(31,73,125)">the </span>node has old replica? Is third small node can
help with it?<span></span></span></p>
</div></div>
</blockquote></div><br><br>-- <br>Sent from Gmail Mobile<br>