[ZOOKEEPER-2082] Mistype of electionAlgo can fill out your disk in minutes

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 3.4.6
• Component/s: leaderElection

Environment

Cluster (multi-server) setup

• Assignee: Unassigned
• Reporter: Tianyin Xu
• Created: 14/Nov/14 22:42
• Updated: 21/Oct/18 07:28
• Resolved: 21/Oct/18 07:28

Description

The parameter, electionAlgo, is supposed to be 0--3. However, when I mistyped the value in my zoo.cfg (I'm stupid), ZK falls into a dead loop and starts filling up the entire disk which millions of the follow 2 lines...

2014-11-14 14:28:44,588 [myid:3] - INFO [QuorumPeer[myid=3]/0:0:0:0:0:0:0:0:2183:QuorumPeer@714] - LOOKING
2014-11-14 14:28:44,588 [myid:3] - WARN [QuorumPeer[myid=3]/0:0:0:0:0:0:0:0:2183:QuorumPeer@764] - Unexpected exception
java.lang.NullPointerException
at org.apache.zookeeper.server.quorum.QuorumPeer.run(QuorumPeer.java:762)

The error rooted in createElectionAlgorithm() where an invalid setting leads to null for the Election object. Then, in the while look in run(), it causes null-pointer de-referencing which is captured but is not handled well.

I think our should check the setting of electionAlg in the very beginning to make sure it's a valid setting, instead of using it at runtime and cause the unfortunate things.

Let me know if you wanna a patch. I'd like to check it in the parseProperties() function in QuorumPeerConfig.java.

Thanks!