[HBASE-10312] Flooding the cluster with administrative actions leads to collapse

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 0.99.0, 0.94.19, 0.98.2, 0.96.3
• Hadoop Flags: Reviewed
• Assignee: Jean-Daniel Cryans
• Reporter: Andrew Kyle Purtell
• Created: 10/Jan/14 19:25
• Updated: 21/Feb/17 04:15
• Resolved: 15/Apr/14 22:54

Description

Steps to reproduce:
1. Start a cluster.
2. Start an ingest process.
3. In the HBase shell, do this:

while true do
   flush 'table'
end

We should reject abuse via administrative requests like this.

What happens on the cluster is the requests back up, leading to lots of these:

2014-01-10 18:55:55,293 WARN  [Priority.RpcServer.handler=2,port=8120] monitoring.TaskMonitor: Too many actions in action monitor! Purging some.

At this point we could lower a gate on further requests for actions until the backlog clears.

Continuing, all of the regionservers will eventually die with a StackOverflowError of unknown origin because, stack overflow:

2014-01-10 19:02:02,783 ERROR [Priority.RpcServer.handler=3,port=8120] ipc.RpcServer: Unexpected throwable object java.lang.StackOverflowError
        at java.util.ArrayList$SubList.add(ArrayList.java:965)
[...]

Issue Links

• breaks: HBASE-17673 Monitored RPC Handler not shown in the WebUI