[ZOOKEEPER-4332] Cannot access children of znode that owns too many znodes

• Type: Bug
• Status: Open
• Priority: Critical
• Resolution: Unresolved
• Affects Version/s: 3.6.1
• Component/s: server
• Labels: pull-request-available zookeeper
• Assignee: Unassigned
• Reporter: Emil Kleszcz
• Created: 30/Jun/21 14:59
• Updated: 27/Jul/21 02:20

Description

We experience problems with performing any operation (deleteall, get etc.) on a znode that has too many child nodes. In our case, it's above 200k. At the same time jute.max.buffer is 4194304. Increasing it by a few factors doesn't help. This should be either solved by limiting the number of direct znodes allowed by a parameter or by adding a hard limit by default.

I am attaching some screenshots of the commands and their results. What's interesting the numbers from getAllChildrenNumber and stat (numChildren) commands don't match.

Issue Links

• duplicates: ZOOKEEPER-1162 consistent handling of jute.maxbuffer when attempting to read large zk "directories"
• relates to: ZOOKEEPER-4314 Can not get real exception when getChildren more than 4M
• links to: GitHub Pull Request #1721