<div dir="ltr"><div dir="ltr"><div>I'm wondering if anyone can advise us on next steps here and/or correct our understanding. This seems like a race condition that causes resources to be stopped unnecessarily. Is there a way to prevent a node from processing cib updates from a peer while DC negotiations are underway? Our "node02" is running resources fine, and since it winds up winning the DC election, would continue to run them uninterrupted if it just ignored or pended the cib updates it receives in the middle of the negotiation.</div><div>Very much appreciate all the help and discussion available on this board.<br></div><div>Regards,</div><div>Stuart</div></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Feb 1, 2021 at 11:43 AM Stuart Massey <<a href="mailto:djangoschef@gmail.com">djangoschef@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-style:solid;border-left-color:rgb(204,204,204);padding-left:1ex"><div dir="ltr">Sequence seems to be:<div><ul><li>node02 is DC and master/primary, node01 is maintenance mode and slave/secondary</li><li>comms go down</li><li>node01 elects itself master, and deletes node01 status from its cib</li><li>comms come up</li><li>cluster starts reforming</li><li>node01 sends cib updates to node02</li><li>DC negotiations start, both nodes unset DC</li><li>node02 receives cib updates and process them, deleting its own status</li><li>DC negotiations complete with node02 winning</li><li>node02, having lost it's status, believes it cannot host resources and stops them all</li><li>for whatever reason, perhaps somehow due to the completely missing transient_attributes, node02 nevers schedules a probe for itself</li><li>we have to "refresh" manually</li></ul></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Feb 1, 2021 at 11:31 AM Ken Gaillot <<a href="mailto:kgaillot@redhat.com" target="_blank">kgaillot@redhat.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-style:solid;border-left-color:rgb(204,204,204);padding-left:1ex">On Mon, 2021-02-01 at 11:09 -0500, Stuart Massey wrote:<br>
> Hi Ken,<br>
> Thanks. In this case, transient_attributes for node02 in the cib on<br>
> node02 which never lost quorum seem to be deleted by a request from<br>
> node01 when node01 rejoins the cluster - IF I understand the<br>
> pacemaker.log correctly. This causes node02 to stop resources, which<br>
> will not be restarted until we manually refresh on node02.<br>
<br>
Good point, it depends on which node is DC. When a cluster splits, each<br>
side sees the other side as the one that left. When the split heals,<br>
whichever side has the newly elected DC is the one that clears the<br>
other.<br>
<br>
However the DC should schedule probes for the other side, and probes<br>
generally set the promotion score, so manual intervention shouldn't be<br>
needed. I'd make sure that probes were scheduled, then investigate how<br>
the agent sets the score.<br>
<br>
> On Mon, Feb 1, 2021 at 10:59 AM Ken Gaillot <<a href="mailto:kgaillot@redhat.com" target="_blank">kgaillot@redhat.com</a>><br>
> wrote:<br>
> > On Fri, 2021-01-29 at 12:37 -0500, Stuart Massey wrote:<br>
> > > Can someone help me with this?<br>
> > > Background:<br>
> > > > "node01" is failing, and has been placed in "maintenance" mode.<br>
> > It<br>
> > > > occasionally loses connectivity.<br>
> > > > "node02" is able to run our resources<br>
> > > <br>
> > > Consider the following messages from pacemaker.log on "node02",<br>
> > just<br>
> > > after "node01" has rejoined the cluster (per "node02"):<br>
> > > > Jan 28 14:48:03 [21933] <a href="http://node02.example.com" rel="noreferrer" target="_blank">node02.example.com</a> cib: <br>
> > info:<br>
> > > > cib_perform_op: --<br>
> > > > /cib/status/node_state[@id='2']/transient_attributes[@id='2']<br>
> > > > Jan 28 14:48:03 [21933] <a href="http://node02.example.com" rel="noreferrer" target="_blank">node02.example.com</a> cib: <br>
> > info:<br>
> > > > cib_perform_op: + /cib: @num_updates=309<br>
> > > > Jan 28 14:48:03 [21933] <a href="http://node02.example.com" rel="noreferrer" target="_blank">node02.example.com</a> cib: <br>
> > info:<br>
> > > > cib_process_request: Completed cib_delete operation for<br>
> > section<br>
> > > > //node_state[@uname='<a href="http://node02.example.com" rel="noreferrer" target="_blank">node02.example.com</a>']/transient_attributes:<br>
> > OK<br>
> > > > (rc=0, origin=<a href="http://node01.example.com/crmd/3784" rel="noreferrer" target="_blank">node01.example.com/crmd/3784</a>, version=0.94.309)<br>
> > > > Jan 28 14:48:04 [21938] <a href="http://node02.example.com" rel="noreferrer" target="_blank">node02.example.com</a> crmd: <br>
> > info:<br>
> > > > abort_transition_graph: Transition aborted by deletion of<br>
> > > > transient_attributes[@id='2']: Transient attribute change |<br>
> > > > cib=0.94.309 source=abort_unless_down:357<br>
> > > ><br>
> > path=/cib/status/node_state[@id='2']/transient_attributes[@id='2']<br>
> > > > complete=true<br>
> > > > Jan 28 14:48:05 [21937] <a href="http://node02.example.com" rel="noreferrer" target="_blank">node02.example.com</a> pengine: <br>
> > info:<br>
> > > > master_color: ms_drbd_ourApp: Promoted 0 instances of a<br>
> > possible 1<br>
> > > > to master<br>
> > > > <br>
> > > The implication, it seems to me, is that "node01" has asked<br>
> > "node02"<br>
> > > to delete the transient-attributes for "node02". The transient-<br>
> > > attributes should normally be:<br>
> > > <transient_attributes id="2"><br>
> > > <instance_attributes id="status-2"><br>
> > > <nvpair id="status-2-master-drbd_ourApp" name="master-<br>
> > > drbd_ourApp" value="10000"/><br>
> > > <nvpair id="status-2-pingd" name="pingd" value="100"/><br>
> > > </instance_attributes><br>
> > > </transient_attributes><br>
> > > <br>
> > > These attributes are necessary for "node02" to be Master/Primary,<br>
> > > correct? <br>
> > > <br>
> > > Why might this be happening and how do we prevent it?<br>
> > <br>
> > Transient attributes are always cleared when a node leaves the<br>
> > cluster<br>
> > (that's what makes them transient ...). It's probably coincidence<br>
> > it<br>
> > went through as the node rejoined.<br>
> > <br>
> > When the node rejoins, it will trigger another run of the<br>
> > scheduler,<br>
> > which will schedule a probe of all resources on the node. Those<br>
> > probes<br>
> > should reset the promotion score.<br>
> > _______________________________________________<br>
> > Manage your subscription:<br>
> > <a href="https://lists.clusterlabs.org/mailman/listinfo/users" rel="noreferrer" target="_blank">https://lists.clusterlabs.org/mailman/listinfo/users</a><br>
> > <br>
> > ClusterLabs home: <a href="https://www.clusterlabs.org/" rel="noreferrer" target="_blank">https://www.clusterlabs.org/</a><br>
-- <br>
Ken Gaillot <<a href="mailto:kgaillot@redhat.com" target="_blank">kgaillot@redhat.com</a>><br>
<br>
_______________________________________________<br>
Manage your subscription:<br>
<a href="https://lists.clusterlabs.org/mailman/listinfo/users" rel="noreferrer" target="_blank">https://lists.clusterlabs.org/mailman/listinfo/users</a><br>
<br>
ClusterLabs home: <a href="https://www.clusterlabs.org/" rel="noreferrer" target="_blank">https://www.clusterlabs.org/</a><br>
</blockquote></div>
</blockquote></div>