<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=ISO-8859-1">
</head>
<body bgcolor="#ffffff" text="#000000">
Hi,<br>
<br>
I'm just migrating my recent mon/heartbeat configuration to pacemaker.
The point of interest is the webservice behavior. Before the monitor
checked if the service failed twice within 20 sec, switch to other node
was initiated if this happens. Now I'm trying to configuring the same
behavior using pacemaker. The webservice is monitored every 10 seconds
(interval=10), failure timeout is set to 20s (expecting to ignore all
failures within this time frame) and it should only happen if a "valid
failure" occurs twice (migration-theshold=2). Valid-failure means: the
service fails twice within 20s but is ignored if the service is back
within 20s. This is the configuration, which is used to implement this
behavior:<br>
<br>
node lbv01 \<br>
attributes standby="off"<br>
node lbv02 \<br>
attributes standby="off"<br>
primitive apacheIP ocf:heartbeat:IPaddr2 \<br>
params ip="10.6.151.190" \<br>
op monitor interval="10s" \<br>
meta is-managed="true"<br>
primitive haproxyIP ocf:heartbeat:IPaddr2 \<br>
params ip="10.6.151.191" \<br>
op monitor interval="10s"<br>
primitive pingd ocf:pacemaker:ping \<br>
params host_list="10.6.151.11" multiplier="100" \<br>
op monitor interval="15s" timeout="5s"<br>
<b>primitive webservice ocf:heartbeat:webservices \<br>
op monitor on-fail="ignore" interval="10s" \<br>
meta failure-timeout="20s" migration-threshold="2"</b><br>
group webservice-ips haproxyIP apacheIP webservice \<br>
meta target-role="Started"<br>
colocation all-resources inf: webservice-ips pingd<br>
property $id="cib-bootstrap-options" \<br>
dc-version="1.1.2-f059ec7ced7a86f18e5490b67ebf4a0b963bccfe" \<br>
cluster-infrastructure="openais" \<br>
expected-quorum-votes="2" \<br>
stonith-enabled="false" \<br>
no-quorum-policy="ignore" \<br>
last-lrm-refresh="1297249441" \<br>
cluster-delay="30"<br>
<br>
If a webservice monitoring failure is forced, the switchover
immediately is performed, ignoring timeout and threshold. How would I
change the configuration to get the expected behavior?<br>
<br>
regards Uwe<br>
<br>
</body>
</html>