Colin McIntosh created ZOOKEEPER-4372:
-----------------------------------------
Summary: Zookeeper lock recipe doesn't parse negative sequence
numbers correctly
Key: ZOOKEEPER-4372
URL: https://issues.apache.org/jira/browse/ZOOKEEPER-4372
Project: ZooKeeper
Issue Type: Bug
Components: recipes
Reporter: Colin McIntosh
According to the Zookeeper docs ZNode sequence numbers will rollover to
-2147483648 after reaching 2147483647 (I've also confirmed this is the case in
practice):
[https://zookeeper.apache.org/doc/r3.7.0/zookeeperProgrammers.html#Sequence+Nodes+–+Unique+Naming|https://zookeeper.apache.org/doc/r3.7.0/zookeeperProgrammers.html#Sequence+Nodes+%E2%80%93+Unique+Naming]
When this rollover happens the negative symbol is incorrectly stripped off the
sequence number in ZNodeName due to this code searching for the last index of
"-", thus the sequence number parsed is always the absolute value:
[https://github.com/apache/zookeeper/blob/1e74a3395a9c3c4d67093f9a073bffb5a176916c/zookeeper-recipes/zookeeper-recipes-lock/src/main/java/org/apache/zookeeper/recipes/lock/ZNodeName.java#L54]
This results in any subsequently created sequence numbers being considered the
"lowest" due to the absolute value decrementing until reaching 0. It also looks
like there is no test to check parsing of negative sequence numbers.
For context, this same issue arose in the unofficial Golang client recipe and
there is a proposed fix there: [https://github.com/go-zookeeper/zk/pull/65]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
