Kirk Lund created GEODE-5345:
--------------------------------
Summary: Region initialization latches should not be invoked twice
when creating a region
Key: GEODE-5345
URL: https://issues.apache.org/jira/browse/GEODE-5345
Project: Geode
Issue Type: Bug
Components: regions
Reporter: Kirk Lund
Log statements reveal that the methods which release initialization latches
(before-initial-image and after-initial-image) are being invoked twice when
creating BucketRegions:
{noformat}
11365-1: [fine 2018/06/21 15:09:27.699 PDT
gemfire1_rs-GEM-1722-klunda0i3large-hydra-client-10_11365 <Pooled Waiting
Message Processor 5> tid=0xd0] Releasing Initialization Latch (before initial
image) for /__PR/_B__partitionedRegion_60
11365-1: [info 2018/06/21 15:09:27.699 PDT
gemfire1_rs-GEM-1722-klunda0i3large-hydra-client-10_11365 <Pooled Waiting
Message Processor 5> tid=0xd0] Region /__PR/_B__partit ionedRegion_60
was created on this member with the persistent id
rs-GEM-1722-klunda0i3large-hydra-client-10/10.32.108.22:/var/vcap/data/rundir/concParRegPersistVOOH4.bt015
/concParRegPersist-0621-150824/vm_0_client1_disk_1 created at timestamp
1529618948528 version 0 diskStoreId efc74f8db5024abb-81bedb7d3f6e038f name null.
11365-1: [fine 2018/06/21 15:09:27.699 PDT
gemfire1_rs-GEM-1722-klunda0i3large-hydra-client-10_11365 <Pooled Waiting
Message Processor 5> tid=0xd0] Releasing Initialization Latch (after initial
image) for /__PR/_B__partitionedRegion_60
1365-1: [info 2018/06/21 15:09:27.699 PDT
gemfire1_rs-GEM-1722-klunda0i3large-hydra-client-10_11365 <Pooled Waiting
Message Processor 5> tid=0xd0] Initialization of region
_B__partitionedRegion_60 completed
11365-1: [fine 2018/06/21 15:09:27.700 PDT
gemfire1_rs-GEM-1722-klunda0i3large-hydra-client-10_11365 <Pooled Waiting
Message Processor 5> tid=0xd0] Releasing Initialization Latch (before initial
image) for /__PR/_B__partitionedRegion_60
11365-1: [fine 2018/06/21 15:09:27.700 PDT
gemfire1_rs-GEM-1722-klunda0i3large-hydra-client-10_11365 <Pooled Waiting
Message Processor 5> tid=0xd0] Releasing Initialization Latch (after initial
image) for /__PR/_B__partitionedRegion_60
{noformat}
These log statements are only logged in one method each:
{noformat}
void releaseBeforeGetInitialImageLatch() {
if (logger.isDebugEnabled()) {
logger.debug("Releasing Initialization Latch (before initial image) for
{}", getFullPath());
}
releaseLatch(this.getInitializationLatchBeforeGetInitialImage());
}
void releaseAfterGetInitialImageLatch() {
if (logger.isDebugEnabled()) {
logger.debug("Releasing Initialization Latch (after initial image) for
{}", getFullPath());
}
releaseLatch(this.getInitializationLatchAfterGetInitialImage());
}
{noformat}
Call hierarchy for these methods shows that there are multiple code paths. I'm
not sure if they are double invoked for non-Bucket Region types or not:
{noformat}
LocalRegion.releaseBeforeGetInitialImageLatch()
(org.apache.geode.internal.cache)
DistributedRegion.getInitialImageAndRecovery(InputStream,
InternalDistributedMember, InternalRegionArguments, boolean,
PersistentMemberID) (org.apache.geode.internal.cache)
DistributedRegion.initialize(InputStream, InternalDistributedMember,
InternalRegionArguments) (org.apache.geode.internal.cache)
LocalRegion.initialize(InputStream, InternalDistributedMember,
InternalRegionArguments) (org.apache.geode.internal.cache)
LocalRegion.releaseLatches() (org.apache.geode.internal.cache)
PartitionedRegion.initPRInternals(InternalRegionArguments)
(org.apache.geode.internal.cache)
{noformat}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
