This is an automated email from the ASF dual-hosted git repository.
zhouxj pushed a commit to branch support/1.14
in repository https://gitbox.apache.org/repos/asf/geode.git
The following commit(s) were added to refs/heads/support/1.14 by this push:
new c9bb764 GEODE-9060: Remove the member from a copy of replicates as
GII candid… (#6246) (#7230)
c9bb764 is described below
commit c9bb7644f210ed6d5a1f35de7dca45659e904a3c
Author: Xiaojian Zhou <[email protected]>
AuthorDate: Mon Jan 3 16:04:18 2022 -0800
GEODE-9060: Remove the member from a copy of replicates as GII candid…
(#6246) (#7230)
* GEODE-9060: Remove the member from a copy of replicates as GII candidate
if
it's not part of the same distributed system, but leave original replicates
unchanged.
(cherry picked from commit 76a5afddc9eaecea9f0b4528a910eb6761e1a3a1)
---
.../PersistentRecoveryOrderDUnitTest.java | 34 +++++++++++++++++++++-
.../cache/persistence/PersistenceAdvisorImpl.java | 18 ++++++++++--
2 files changed, 48 insertions(+), 4 deletions(-)
diff --git
a/geode-core/src/distributedTest/java/org/apache/geode/internal/cache/persistence/PersistentRecoveryOrderDUnitTest.java
b/geode-core/src/distributedTest/java/org/apache/geode/internal/cache/persistence/PersistentRecoveryOrderDUnitTest.java
index 1270e89..4f1f436 100644
---
a/geode-core/src/distributedTest/java/org/apache/geode/internal/cache/persistence/PersistentRecoveryOrderDUnitTest.java
+++
b/geode-core/src/distributedTest/java/org/apache/geode/internal/cache/persistence/PersistentRecoveryOrderDUnitTest.java
@@ -876,11 +876,43 @@ public class PersistentRecoveryOrderDUnitTest extends
CacheTestCase {
Throwable thrown = catchThrowable(() -> {
createReplicateRegion(regionName, getDiskDirs(getVMId()));
});
-
assertThat(thrown).isInstanceOf(ConflictingPersistentDataException.class);
+ assertThat(thrown)
+ .isInstanceOf(ConflictingPersistentDataException.class)
+ .hasMessageContaining("was not part of the same distributed system
as the local data");
}
});
}
+ @Test
+ public void testRecoverableSplitBrain() {
+ vm2.invoke(() -> {
+ createReplicateRegion(regionName, getDiskDirs(getVMId()));
+ });
+ vm0.invoke(() -> {
+ createReplicateRegion(regionName, getDiskDirs(getVMId()));
+ putEntry("A", "B");
+ getCache().getRegion(regionName).close();
+ });
+
+ vm1.invoke(() -> {
+ createReplicateRegion(regionName, getDiskDirs(getVMId()));
+ validateEntry("A", "B");
+ updateEntry("A", "C");
+ getCache().getRegion(regionName).close();
+ });
+
+ // VM0 doesn't know that VM1 ever existed so it will start up.
+ vm0.invoke(() -> {
+ createReplicateRegion(regionName, getDiskDirs(getVMId()));
+ validateEntry("A", "C");
+ });
+
+ vm1.invoke(() -> {
+ createReplicateRegion(regionName, getDiskDirs(getVMId()));
+ validateEntry("A", "C");
+ });
+ }
+
/**
* Test to make sure that if if a member crashes while a GII is in progress,
we wait for the
* member to come back for starting.
diff --git
a/geode-core/src/main/java/org/apache/geode/internal/cache/persistence/PersistenceAdvisorImpl.java
b/geode-core/src/main/java/org/apache/geode/internal/cache/persistence/PersistenceAdvisorImpl.java
index 11f2563..073a632 100644
---
a/geode-core/src/main/java/org/apache/geode/internal/cache/persistence/PersistenceAdvisorImpl.java
+++
b/geode-core/src/main/java/org/apache/geode/internal/cache/persistence/PersistenceAdvisorImpl.java
@@ -510,18 +510,18 @@ public class PersistenceAdvisorImpl implements
InternalPersistenceAdvisor {
public boolean checkMyStateOnMembers(Set<InternalDistributedMember>
replicates)
throws ReplyException {
PersistentStateQueryResults remoteStates = getMyStateOnMembers(replicates);
+ Set<InternalDistributedMember> copyOfReplicates = null;
persistenceAdvisorObserver.observe(regionPath);
boolean equal = false;
+ PersistentMemberID myId = getPersistentID();
for (Map.Entry<InternalDistributedMember, PersistentMemberState> entry :
remoteStates
.getStateOnPeers().entrySet()) {
InternalDistributedMember member = entry.getKey();
PersistentMemberID remoteId =
remoteStates.getPersistentIds().get(member);
- PersistentMemberID myId = getPersistentID();
PersistentMemberState stateOnPeer = entry.getValue();
-
if (PersistentMemberState.REVOKED.equals(stateOnPeer)) {
throw new RevokedPersistentDataException(
String.format(
@@ -533,7 +533,19 @@ public class PersistenceAdvisorImpl implements
InternalPersistenceAdvisor {
String message = String.format(
"Region %s remote member %s with persistent data %s was not part
of the same distributed system as the local data from %s",
regionPath, member, remoteId, myId);
- throw new ConflictingPersistentDataException(message);
+ // Conceptually the removed member due to not knowing current member,
should be equal to
+ // existing replicates.
+ // It can still be used as GII provider candidate. Use a
copyOfReplicates to avoid modifying
+ // the replicates.
+ if (copyOfReplicates == null) {
+ copyOfReplicates = new HashSet<>(replicates);
+ }
+ copyOfReplicates.remove(member);
+ if (copyOfReplicates.isEmpty()) {
+ throw new ConflictingPersistentDataException(message);
+ } else {
+ logger.info(message);
+ }
}
if (myId != null && stateOnPeer == PersistentMemberState.EQUAL) {
