[
https://issues.apache.org/jira/browse/HBASE-1394?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12707845#action_12707845
]
stack commented on HBASE-1394:
------------------------------
Looking at log sizes, they are 110MB regularly in this test. So, there is at
least one block transition inline. That probably costs.
So, here are some tests to try using HLog and HDFS only -- can be done without
involving regionserver -- just write a harness:
1. See if writing logs < hdfs block size improves things
2. In my measurements, it doesn't seem to be an issue but close of a log file
during roll is inline -- all writes stop while the close is being done.
Measure doing close out of band. Measure too opening the new log file before
we need it so that its read to go rather than hold up incoming edits while new
file is opened.
I don't think the above will address the long time I'm seeing doing appends at
certain times during a bulk upload. Talking with Ryan, maybe we need a pool of
loggers where we scatter edits over the pool -- if one is off waiting on a
DataStreamer to catchup then that one write will take some time rather than
have all blocked. This'll make the recovery harder in that we'll have to sort
regionservers's log pool but we're going to have to rewrite that anyways.
> Uploads sometimes fall to 0 requests/second (Binding up on HLog#append?)
> ------------------------------------------------------------------------
>
> Key: HBASE-1394
> URL: https://issues.apache.org/jira/browse/HBASE-1394
> Project: Hadoop HBase
> Issue Type: Bug
> Reporter: stack
> Attachments: loghowlongittookappendinghlog.patch
>
>
> Trying to figure why rate sometimes goes to zero.
> Studying the reginoserver, HLog#append looks like a possible culprit.
> {code}
> "IPC Server handler 7 on 60021" daemon prio=10 tid=0x000000004057dc00
> nid=0x1bc4 in Object.wait() [0x0000000043393000..0x0000000043393b80]
> java.lang.Thread.State: WAITING (on object monitor)
> at java.lang.Object.wait(Native Method)
> at java.lang.Object.wait(Object.java:485)
> at
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.writeChunk(DFSClient.java:2964)
> - locked <0x00007f9e3e449ff0> (a java.util.LinkedList)
> - locked <0x00007f9e3e449e18> (a
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream)
> at
> org.apache.hadoop.fs.FSOutputSummer.writeChecksumChunk(FSOutputSummer.java:150)
> at org.apache.hadoop.fs.FSOutputSummer.write1(FSOutputSummer.java:100)
> at org.apache.hadoop.fs.FSOutputSummer.write(FSOutputSummer.java:86)
> - locked <0x00007f9e3e449e18> (a
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream)
> at
> org.apache.hadoop.fs.FSDataOutputStream$PositionCache.write(FSDataOutputStream.java:49)
> at java.io.DataOutputStream.write(DataOutputStream.java:90)
> - locked <0x00007f9e434e5588> (a
> org.apache.hadoop.fs.FSDataOutputStream)
> at
> org.apache.hadoop.io.SequenceFile$Writer.append(SequenceFile.java:1020)
> - locked <0x00007f9e434e55c0> (a
> org.apache.hadoop.io.SequenceFile$Writer)
> at
> org.apache.hadoop.io.SequenceFile$Writer.append(SequenceFile.java:984)
> - locked <0x00007f9e434e55c0> (a
> org.apache.hadoop.io.SequenceFile$Writer)
> at org.apache.hadoop.hbase.regionserver.HLog.doWrite(HLog.java:565)
> at org.apache.hadoop.hbase.regionserver.HLog.append(HLog.java:521)
> - locked <0x00007f9dfa376f70> (a java.lang.Object)
> at
> org.apache.hadoop.hbase.regionserver.HRegion.update(HRegion.java:1777)
> at
> org.apache.hadoop.hbase.regionserver.HRegion.batchUpdate(HRegion.java:1348)
> at
> org.apache.hadoop.hbase.regionserver.HRegion.batchUpdate(HRegion.java:1289)
> at
> org.apache.hadoop.hbase.regionserver.HRegionServer.batchUpdates(HRegionServer.java:1727)
> at sun.reflect.GeneratedMethodAccessor9.invoke(Unknown Source)
> at
> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
> at java.lang.reflect.Method.invoke(Method.java:597)
> at org.apache.hadoop.hbase.ipc.HBaseRPC$Server.call(HBaseRPC.java:642)
> at
> org.apache.hadoop.hbase.ipc.HBaseServer$Handler.run(HBaseServer.java:911)
> {code}
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
