KernelHealth is a component that keeps state of the last kernel panic and prohibits database writes until it is resolved. So far this state was never reset even though it could happen that internal restarts (in the case of HA) might lead to recovery and resolution of the panic cause. This patch makes sure that after a successful startup cycle on NSDS the kernel health is restored.
Setting up a NeoStoreDataSource for testing can be complicated. This patch introduces a JUnit @Rule for that purpose and uses it in TestNeoStore
… step of import tool. Should only occur on really small datasets (< 5) however.
Previously the errors were reporting the wrong position due to the fact that part of the string input was consumed by a pre-parser, now we make sure to consider the pre-parsed offset when reporting errors.
Conflicts: community/kernel/src/test/java/org/neo4j/kernel/impl/core/TestCrashWithRebuildSlow.java community/kernel/src/test/java/org/neo4j/kernel/impl/nioneo/store/TestBrokenStoreRecovery.java community/kernel/src/test/java/org/neo4j/kernel/impl/transaction/xaframework/TestApplyTransactions.java community/kernel/src/test/java/org/neo4j/kernel/impl/transaction/xaframework/TestTxEntries.java community/neo4j/src/test/java/recovery/TestRecoveryScenarios.java
Made this a null-erge, nothing changed. I have no idea how to implement this for 2.2, it'll have to a re-implementation of the original author. Conflicts: community/kernel/src/main/java/org/neo4j/kernel/impl/locking/community/CommunityLockClient.java community/kernel/src/main/java/org/neo4j/kernel/impl/locking/community/LockTransaction.java community/kernel/src/main/java/org/neo4j/kernel/impl/transaction/TransactionImpl.java community/kernel/src/main/java/org/neo4j/kernel/impl/transaction/TransactionStateFactory.java community/kernel/src/test/java/org/neo4j/kernel/impl/transaction/TxManagerTest.java enterprise/ha/src/main/java/org/neo4j/kernel/ha/lock/SlaveLocksClient.java
…isualVM, thus making it impossible to GC the entire database. This causes OOMs during testing when we create/destroy databases within one JVM session. Fix is to clear all references in STM on shutdown.
…ties this would happen on batches where there were skipped relationships. Any relationships after skipped relationships in that batch would have their sets of properties shifted the number of steps there were skipped relationships at that position. So the import would create a consistent database, just semantically incorrect.
since that's just overhead. Querying will (or at least is able to) handle absence of such values anyway.
- Add text to hint the user why there's nothing to show. It could be that nothing was returned or the returned results were empty. - Add `(empty)` holder to show that a returned node has no properties OR that a returned string has no length (it exists, but is empty).