[SOLR-16877] solr backup via collections api fails with NPE when config contains an empty zookeeper node

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 9.2.1
• Fix Version/s: 9.3
• Component/s: Backup/Restore
• Labels: backup collection-api
• Assignee: Houston Putman
• Reporter: Paul Blanchaert
• Created: 07/Jul/23 13:17
• Updated: 21/Jul/23 20:42
• Resolved: 10/Jul/23 15:26

Description

The root cause for the backup to fail is an empty node in zookeeper (E.g. the querqy request handler in solr creates an empty .data node when starting up).

To reproduce the issue, create a new (empty) node in zk (via zkCli.sh session and "create" command) under a configset of your collection and back up this collection via the collections API with action=BACKUP command.

The reason is that the byte array provided to java.io.OutputStream.write is null in the case of an empty zookeeper node.

Error:
2023-07-07 12:33:52.121 DEBUG (OverseerThreadFactory-17-thread-3) [c:gettingstarted_20230707] o.a.s.c.b.BackupManager Writing file querqy/rewriters/.data
2023-07-07 12:33:52.123 ERROR (OverseerThreadFactory-17-thread-3) [c:gettingstarted_20230707] o.a.s.c.a.c.OverseerCollectionMessageHandler Collection: gettingstarted_20230707 operation: backup failed => java.lang.NullPointerException: Cannot read the array length because "<parameter1>" is null
at java.base/java.io.OutputStream.write(Unknown Source)
java.lang.NullPointerException: Cannot read the array length because "<parameter1>" is null
at java.io.OutputStream.write(Unknown Source) ~[?:?]
at org.apache.solr.core.backup.BackupManager.downloadConfigToRepo(BackupManager.java:349) ~[?:?]
at org.apache.solr.core.backup.BackupManager.downloadConfigDir(BackupManager.java:276) ~[?:?]
at org.apache.solr.cloud.api.collections.BackupCmd.call(BackupCmd.java:149) ~[?:?]
at org.apache.solr.cloud.api.collections.OverseerCollectionMessageHandler.processMessage(OverseerCollectionMessageHandler.java:126) ~[?:?]
at org.apache.solr.cloud.OverseerTaskProcessor$Runner.run(OverseerTaskProcessor.java:566) ~[?:?]
at org.apache.solr.common.util.ExecutorUtil$MDCAwareThreadPoolExecutor.lambda$execute$0(ExecutorUtil.java:289) ~[?:?]
at java.util.concurrent.ThreadPoolExecutor.runWorker(Unknown Source) ~[?:?]
at java.util.concurrent.ThreadPoolExecutor$Worker.run(Unknown Source) ~[?:?]
at java.lang.Thread.run(Unknown Source) [?:?]

 

Note:

• when you create an empty file (with touch) and copy it in zookeeper (with "solr zk cp"), the backup doesn't raise the error (and backup succeeds).
• when you create an empty node in zookeeper (via zkCli.sh session and "create" command), the backup raises the same error.

Issue Links

• links to: GitHub Pull Request #1764