[BOOKKEEPER-195] HierarchicalLedgerManager doesn't consider idgen as a "specialNode"

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 4.1.0
• Assignee: Ivan Kelly
• Reporter: Ivan Kelly
• Created: 28/Mar/12 20:10
• Updated: 22/Oct/12 14:50
• Resolved: 28/Mar/12 22:02

Description

From IRC:

21:58 <johnnagro> 2012-03-28 05:15:19,669 - ERROR [main-EventThread:AbstractZkLedgerManager$2@105] - Error polling ZK for the available ledger nodes:
21:58 <johnnagro> org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode = NoNode for /ledgers/idgen/ID-0000106522
21:58 <johnnagro> at org.apache.zookeeper.KeeperException.create(KeeperException.java:111)
21:58 <johnnagro> at org.apache.zookeeper.KeeperException.create(KeeperException.java:51)
21:58 <johnnagro> at org.apache.bookkeeper.meta.AbstractZkLedgerManager$2.processResult(AbstractZkLedgerManager.java:105)
21:58 <johnnagro> at org.apache.zookeeper.ClientCnxn$EventThread.processEvent(ClientCnxn.java:594)
21:58 <johnnagro> at org.apache.zookeeper.ClientCnxn$EventThread.run(ClientCnxn.java:497)
21:58 <johnnagro> i will look for the gc one
21:59 <ivank> thats quite strange
22:00 <ivank> this occurred in the server?
22:02 <johnnagro> ah they may be related, this one has some GC stuff
22:02 <johnnagro> 012-03-24 09:17:34,419 - WARN [GarbageCollectorThread:HierarchicalLedgerManager@376] - Exception during garbage collecting ledgers for idgen of /ledgers
22:02 <johnnagro> 2012-03-24 09:56:31,080 - ERROR [main-EventThread:AbstractZkLedgerManager$2@105] - Error polling ZK for the available ledger nodes:
22:02 <johnnagro> org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode = NoNode for /ledgers/idgen/ID-0000037474
22:02 <johnnagro> at org.apache.zookeeper.KeeperException.create(KeeperException.java:111)
22:02 <johnnagro> at org.apache.zookeeper.KeeperException.create(KeeperException.java:51)
22:02 <johnnagro> at org.apache.bookkeeper.meta.AbstractZkLedgerManager$2.processResult(AbstractZkLedgerManager.java:105)
22:02 <johnnagro> at org.apache.zookeeper.ClientCnxn$EventThread.processEvent(ClientCnxn.java:594)
22:02 <johnnagro> at org.apache.zookeeper.ClientCnxn$EventThread.run(ClientCnxn.java:497)
22:02 <johnnagro> this happened on the server, yes.

Looking into the code, it seems that HierarchicalLedgerManager doesn't consider idgen to be special. This is a race, between the idgen node being created and the znodes underneath being listed. The znodes underneath should never be listed.