Kirk Lund created GEODE-2244:
--------------------------------
Summary: Cluster config start is async and may not be ready when
requests start coming in
Key: GEODE-2244
URL: https://issues.apache.org/jira/browse/GEODE-2244
Project: Geode
Issue Type: Bug
Reporter: Kirk Lund
This wouldn't be a problem but it looks like the locator hosting cluster config
may reply to requests saying that it doesn't have cluster config enabled.
The method in InternalLocator is startCache(DistributedSystem):
{code:java}
private void startCache(DistributedSystem ds) {
GemFireCacheImpl gfc = GemFireCacheImpl.getInstance();
if (gfc == null) {
logger.info("Creating cache for locator.");
this.myCache = new CacheFactory(ds.getProperties()).create();
gfc = (GemFireCacheImpl) this.myCache;
} else {
logger.info("Using existing cache for locator.");
((InternalDistributedSystem)
ds).handleResourceEvent(ResourceEvent.LOCATOR_START, this);
}
startJmxManagerLocationService(gfc);
startSharedConfigurationService(gfc);
}
{code}
The method startSharedConfigurationService hands off to a thread to load
cluster config and use distributed lock service to become primary cluster
config source. This was probably made async due to use of distributed lock
service to keep startup responsive. Unfortunately, it opens up a race condition
window -- if any requests come in before cluster config is ready, the locator
will reply saying it doesn't have cluster config. I think some of our Flaky
tests are hitting this race condition.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
