neo4j升级1.9.x-> 2.x失败

Question

I'm trying to run an upgrade from neo4j 1.9.8 to 2.3.0

I'm using neo4j in embedded java mode, with additional webadmin configured via org.neo4j.server.configuration.ServerConfigurator .

I have

allow_store_upgrade=true

And the following appears in messages.log

2015-10-22 09:45:19.263+0000 INFO  [o.n.k.i.s.StoreMigrator] Starting upgrade of database store files
2015-10-22 09:45:19.609+0000 INFO  [o.n.u.i.b.ParallelBatchImporter] Import starting
2015-10-22 09:45:19.662+0000 WARN  [o.n.k.i.s.MetaDataStore] Missing counts store, rebuilding it.
2015-10-22 09:45:19.870+0000 INFO  [o.n.k.a.i.i.LuceneLabelScanStore] No lucene scan store index found, this might just be first use. Preparing to rebuild.
2015-10-22 09:45:20.011+0000 INFO  [o.n.k.a.i.i.LuceneLabelScanStore] Rebuilding lucene scan store, this may take a while
2015-10-22 09:45:20.013+0000 INFO  [o.n.k.a.i.i.LuceneLabelScanStore] Lucene scan store rebuilt (roughly -1 nodes)
2015-10-22 09:47:40.676+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 10% complete
2015-10-22 09:49:30.698+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 20% complete
2015-10-22 09:57:09.822+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 30% complete
2015-10-22 10:00:36.621+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 40% complete
2015-10-22 10:08:16.226+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 50% complete
2015-10-22 10:16:30.454+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 60% complete
2015-10-22 10:20:19.178+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 70% complete
2015-10-22 10:35:07.315+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 80% complete
2015-10-22 10:58:56.521+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 90% complete
2015-10-22 11:11:26.017+0000 INFO  [o.n.k.i.s.StoreMigrator] Store upgrade 100% complete
2015-10-22 11:15:32.419+0000 INFO  [o.n.u.i.b.ParallelBatchImporter] Import completed, took 1h 30m 12s 805ms. Imported:
  427688598 nodes
  884330252 relationships
  0 properties

So it says it's done, but the claim of 0 properties is suspicious.

When I then remove allow_store_upgrade property and re-boot, I get:

Caused by: org.neo4j.kernel.impl.storemigration.UpgradeNotAllowedByConfigurationException: Failed to start Neo4j with an older data store version. To enable automatic upgrade, please set configuration parameter "allow_store_upgrade=true"
        at org.neo4j.kernel.impl.storemigration.ConfigMapUpgradeConfiguration.checkConfigurationAllowsAutomaticUpgrade(ConfigMapUpgradeConfiguration.java:39) ~[neo4j-kernel-2.3.0-RC1.jar:2.3.0-RC1,8014933]
        at org.neo4j.kernel.impl.storemigration.StoreUpgrader.migrateIfNeeded(StoreUpgrader.java:135) ~[neo4j-kernel-2.3.0-RC1.jar:2.3.0-RC1,8014933]
        at org.neo4j.kernel.NeoStoreDataSource.upgradeStore(NeoStoreDataSource.java:623) ~[neo4j-kernel-2.3.0-RC1.jar:2.3.0-RC1,8014933]
        at org.neo4j.kernel.NeoStoreDataSource.start(NeoStoreDataSource.java:514) ~[neo4j-kernel-2.3.0-RC1.jar:2.3.0-RC1,8014933]
        at org.neo4j.kernel.lifecycle.LifeSupport$LifecycleInstance.start(LifeSupport.java:528) ~[neo4j-kernel-2.3.0-RC1.jar:2.3.0-RC1,8014933]
        ... 89 common frames omitted

Answer 1

Make sure that you have a clean shutdown on your 1.9.x database. To be on the safe side run consistency checker on that one, see http://www.markhneedham.com/blog/2014/01/22/neo4j-backup-store-copy-and-consistency-check/ for details.

My expectation is that 1.9 -> 2.3 upgrade should run in one step. If that doesn't work for you, try to run the upgrade in multiple steps: 1.9 -> 2.0 -> 2.1 -> 2.2 -> 2.3. If that multiple step approach succeeds, I consider the failure of the "large" upgrade a bug worth filing an issue at https://github.com/neo4j/neo4j/issues .

Answer 2

Upgrade directly from 1.9 --> 2.3 is supported. The note about 0 properties is misleading but benign since the migration process doesn't touch the properties, but leave them as-is.

Although it's very weird that it complains about the upgrade not allowed after the initial migration completes (the next startup). If the database wouldn't have been cleanly shut down before the migration, there should have been an exception saying migration wouldn't be possible.