Hao Zhong created CASSANDRA-13693:
-------------------------------------
Summary: A potential problem in the
Ec2MultiRegionSnitch_gossiperStarting method
Key: CASSANDRA-13693
URL: https://issues.apache.org/jira/browse/CASSANDRA-13693
Project: Cassandra
Issue Type: Bug
Reporter: Hao Zhong
The code of Ec2MultiRegionSnitch_gossiperStarting is as follow:
{code}
public void gossiperStarting()
{
super.gossiperStarting();
Gossiper.instance.addLocalApplicationState(ApplicationState.INTERNAL_IP,
StorageService.instance.valueFactory.internalIP(localPrivateAddress));
Gossiper.instance.register(new ReconnectableSnitchHelper(this,
ec2region, true));
}
{code}
I notice that CASSANDRA-5897 fixed a bug, whose buggy code is identical. The
fixed code is
{code}
public void gossiperStarting()
{
super.gossiperStarting();
Gossiper.instance.addLocalApplicationState(ApplicationState.INTERNAL_IP,
StorageService.instance.valueFactory.internalIP(FBUtilities.getLocalAddress().getHostAddress()));
reloadGossiperState();
gossipStarted = true;
}
private void reloadGossiperState()
{
if (Gossiper.instance != null)
{
ReconnectableSnitchHelper pendingHelper = new
ReconnectableSnitchHelper(this, myDC, preferLocal);
Gossiper.instance.register(pendingHelper);
pendingHelper = snitchHelperReference.getAndSet(pendingHelper);
if (pendingHelper != null)
Gossiper.instance.unregister(pendingHelper);
}
// else this will eventually rerun at gossiperStarting()
}
{code}
If Ec2MultiRegionSnitch is supposed to auto-reload, the above fix shall be
applied to its code.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
