hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mike Spreitzer <mspre...@us.ibm.com>
Subject Understanding responseTooSlow in HBase regionserver log
Date Thu, 26 Apr 2012 17:55:52 GMT
I found http://hbase.apache.org/book/ops.monitoring.html but am confused 
about the distinction between operationTooSlow and responseTooSlow.  I 
find the text "client operation" ambiguous, I am not sure whether that 
means the side emitting this log entry is the client and the slow 
operation happened on some server elsewhere or it is the other way around 
or something else.  I am running HBase-0.92.0 and found the following in 
the log of a regionserver:

2012-04-26 13:17:30,698 WARN org.apache.hadoop.ipc.HBaseServer: 
(responseTooSlow): 
{"processingtimems":10340,"call":"execCoprocessor([B@f2d376, 
<mycoprocessor/>, rpc version=1, client version=0, methodsFingerPrint=0), 
rpc version=1, client version=29, 
methodsFingerPrint=54742778","client":"10.12.203.173:58531","starttimems":1335460640355,"queuetimems":0,"class":"HRegionServer","responsesize":0,"method":"execCoprocessor"}

10.12.203.173 is not the IP address of the machine running this 
regionserver, it is the IP address of the machine running my HBase client 
application.

When is operationTooSlow used and when is responseTooSlow used?

In my example above, am I correct that this was merely a warning issued on 
the regionserver saying that my coprocessor took a bit over 10 seconds on 
that regionserver and this did not cause any actual harm?  Note that my 
HBase client application printed no complaints and appeared to work 
correctly (although my evaluation of "correctness" was limited).

Thanks,
Mike
Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message