[ZOOKEEPER-4219] Quota checks break setData in multi transactions

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 3.7.0, 3.8.0
• Fix Version/s: 3.7.0, 3.8.0
• Labels: pull-request-available
• Assignee: Damien Diederen
• Reporter: Damien Diederen
• Created: 23/Feb/21 09:04
• Updated: 28/Mar/21 08:54
• Resolved: 06/Mar/21 19:54

Description

On a server with enforceQuota=true, multi() transactions containing setData operations can throw NullPointerException during the quota check.

On the server side, this looks like:

java.lang.NullPointerException
    at org.apache.zookeeper.server.ZooKeeperServer.checkQuota(ZooKeeperServer.java:2048)
    at org.apache.zookeeper.server.PrepRequestProcessor.pRequest2Txn(PrepRequestProcessor.java:397)
        [...]

On the client side, this is reflected as a MarshallingError:

org.apache.zookeeper.KeeperException$MarshallingErrorException: KeeperErrorCode = MarshallingError
    at org.apache.zookeeper.KeeperException.create(KeeperException.java:104)
    at org.apache.zookeeper.ZooKeeper.multiInternal(ZooKeeper.java:1758)
        [...]

This is visibly due to the quota check trying to access node data by path from the PrepRequestProcessor without considering the collection of outstandingChanges. I have a patch in the works.

Issue Links

• links to: GitHub Pull Request #1611