Mailing-List: contact dev-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hbase.apache.org
Received-SPF: pass (nike.apache.org: local policy)
From: Vladimir Rodionov <vrodionov@carrieriq.com>
To: "dev@hbase.apache.org" <dev@hbase.apache.org>
Date: Wed, 2 Nov 2011 13:24:27 -0700
Subject: HBase 0.90.4 missing data in production
Thread-Topic: HBase 0.90.4 missing data in production
Thread-Index: AQHMmZ1riwGPuBrGDUq6u1MtEvPxXQ==
Message-ID: 
 <DC5EBE7F3610EB4CA5C7E92D76873E15170BE3E374@exchange2007.carrieriq.com>
Accept-Language: en-US
Content-Language: en-US
acceptlanguage: en-US
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0

Even with WAL enabled
HBase version: 0.90.4
Hadoop: 0.20.2+320 (CDH3)


When M/R job finishes loading data into HBase table and right after that ev=
erything is OK, no missing data. We can confirm that by running  our own in=
ternal tool.
But when we re-start the HBase cluster and run the checker again we regular=
ly find missing rows in a table. The total number of rows decreases in a ta=
ble.
This "missing rows" amount is small =3D ~ 0.1-0.3% of a total. After first =
restart all subsequent restart do not affect on a total number of rows. It =
seems,
that we loose only during first restart

Table's TTL =3D 1 year. There is a slim chance that we load data with times=
tamps more than one year behind, but it does not explain the difference
between total number of rows before and after cluster's restart.

All RS are time synched.

In Master log I do not see any WARN or ERRORs during cluster re-start. In R=
S logs I see a lot of:

2011-11-02 00:16:07,620 INFO org.apache.hadoop.hbase.regionserver.wal.HLog:=
 Roll /hbase/.logs/us01-ciqps1-grid01.carrieriq.com,60020,1320187507171/us0=
1-ciqps1-grid01.carrieriq.com%3A60020.1320192949451, entries=3D76, filesize=
=3D68053806. New hlog /hbase/.logs/us01-ciqps1-grid01.carrieriq.com,60020,1=
320187507171/us01-ciqps1-grid01.carrieriq.com%3A60020.1320192967380
2011-11-02 00:16:07,621 INFO org.apache.hadoop.io.compress.CodecPool: Got b=
rand-new compressor
2011-11-02 00:16:07,621 FATAL org.apache.hadoop.hbase.regionserver.wal.HLog=
: Could not append. Requesting close of hlog
java.io.IOException: Reflection
        at org.apache.hadoop.hbase.regionserver.wal.SequenceFileLogWriter.s=
ync(SequenceFileLogWriter.java:147)
        at org.apache.hadoop.hbase.regionserver.wal.HLog.sync(HLog.java:100=
2)
        at org.apache.hadoop.hbase.regionserver.wal.HLog.append(HLog.java:9=
55)
        at org.apache.hadoop.hbase.regionserver.HRegion.doMiniBatchPut(HReg=
ion.java:1483)
        at org.apache.hadoop.hbase.regionserver.HRegion.put(HRegion.java:13=
92)
        at org.apache.hadoop.hbase.regionserver.HRegionServer.multi(HRegion=
Server.java:2591)
        at sun.reflect.GeneratedMethodAccessor13.invoke(Unknown Source)
        at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethod=
AccessorImpl.java:25)
        at java.lang.reflect.Method.invoke(Method.java:597)
        at org.apache.hadoop.hbase.ipc.HBaseRPC$Server.call(HBaseRPC.java:5=
70)
        at org.apache.hadoop.hbase.ipc.HBaseServer$Handler.run(HBaseServer.=
java:1039)
Caused by: java.lang.reflect.InvocationTargetException
        at sun.reflect.GeneratedMethodAccessor2.invoke(Unknown Source)
        at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethod=
AccessorImpl.java:25)
        at java.lang.reflect.Method.invoke(Method.java:597)
        at org.apache.hadoop.hbase.regionserver.wal.SequenceFileLogWriter.s=
ync(SequenceFileLogWriter.java:145)
        ... 10 more
Caused by: java.lang.NullPointerException
        at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.closeThreads(DF=
SClient.java:3306)
        at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.sync(DFSClient.=
java:3216)
        at org.apache.hadoop.fs.FSDataOutputStream.sync(FSDataOutputStream.=
java:97)
        at org.apache.hadoop.io.SequenceFile$Writer.syncFs(SequenceFile.jav=
a:944)
        ... 14 more

This is probably not all ERRORS and FATALs I am continuing investigation an=
d will post my other findings later.

Can someone tell me what should I check else?

Best regards,
Vladimir Rodionov
Principal Platform Engineer
Carrier IQ, www.carrieriq.com
e-mail: vrodionov@carrieriq.com

________________________________________
From: Todd Lipcon [todd@cloudera.com]
Sent: Wednesday, November 02, 2011 12:53 PM
To: dev@hbase.apache.org
Subject: Re: HBASE-4120 table level priority

It's up to each committer to decide how they want to allocate their
time, of course, but I'd like to encourage folks to hold off on
spending a lot of time on new 0.94 feature merges until we've gotten
0.92.0 stabilized and out the door. Big features like QoS need a fair
amount of review, both on the general approach and on the specific
implementation, before they can get merged. It's hard to find time for
a thorough review when we're in the finishing stages of stabilizing a
release.

-Todd

On Wed, Nov 2, 2011 at 12:23 PM, Ted Yu <yuzhihong@gmail.com> wrote:
> Hi,
> HBASE-4120 is a new feature for 0.94 that allows users to set different
> priorities for their tables.
> Priorities can also be assigned to methods (get's vs. put's).
>
> The new tests in the patch all pass.
>
> Please review the latest patch and provide your comments.
>
> Thanks
>


--
Todd Lipcon
Software Engineer, Cloudera

Confidentiality Notice:  The information contained in this message, includi=
ng any attachments hereto, may be confidential and is intended to be read o=
nly by the individual or entity to whom this message is addressed. If the r=
eader of this message is not the intended recipient or an agent or designee=
 of the intended recipient, please note that any review, use, disclosure or=
 distribution of this message or its attachments, in any form, is strictly =
prohibited.  If you have received this message in error, please immediately=
 notify the sender and/or Notifications@carrieriq.com and delete or destroy=
 any copy of this message and its attachments.