Aled Sage created BROOKLYN-145:
----------------------------------
Summary: SameServerEntity reported incorrectly (temporary) "up"
before children started
Key: BROOKLYN-145
URL: https://issues.apache.org/jira/browse/BROOKLYN-145
Project: Brooklyn
Issue Type: Bug
Affects Versions: 0.7.0-SNAPSHOT
Reporter: Aled Sage
The ComputeServiceIndicatorsFromChildrenAndMembers.UP_QUORUM_CHECK behaviour
recently changed: it has default atLeastOneUnlessEmpty, and filters out those
children/members that have not yet started.
The consequence for SameServerEntity is that it incorrectly reports "up"
temporarily at the beginning of startup. When its children were in expected
states (i.e. none were on-fire, and all were still off because start() had not
been called on them), then the SameServerEntity reported "up". This was because
all non-started children where filtered out, so the quorum check treated it as
"empty" (i.e. no children/members).
A customer is using a SameServerEntity to start two processes, and has another
entity configured with an INSTALL_LATCH to block until the SameServerEntity is
up. The temporary "up" was enough to unblock the latch, causing the other
entity to continue prematurely.
I suggest we change the default in SameServerEntity. For example:
{noformat}
@Override
protected void initEnrichers() {
super.initEnrichers();
addEnricher(ServiceStateLogic.newEnricherFromChildren()
.configure(ComputeServiceIndicatorsFromChildrenAndMembers.UP_QUORUM_CHECK,
QuorumCheck.QuorumChecks.allAndAtLeastOne()));
}
{noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
