HBase, mail # user - Re: How to check if a major_compact is done?

How to check if a major_compact is done?

yun peng 2012-11-08, 06:08

Hi, All,
I want to measure the duration of a major compaction in HBase. Since the
function call majorCompact is asynchronous, I may need to manually check
when the major compaction is done. Does Hbase (as of version 0.92.4)
provide an interface to determine completion of major compaction? Thanks.
Yun

Re: How to check if a major_compact is done?

ramkrishna vasudevan 2012-11-08, 06:33

There is no interface which says that the major compaction is completed.
But you can see that major compaction is in progress from the web UI.
Sorry if am wrong here.

Regards
Ram

On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]> wrote:

> Hi, All,
> I want to measure the duration of a major compaction in HBase. Since the
> function call majorCompact is asynchronous, I may need to manually check
> when the major compaction is done. Does Hbase (as of version 0.92.4)
> provide an interface to determine completion of major compaction? Thanks.
> Yun
>

Re: How to check if a major_compact is done?

PG 2012-11-08, 08:08

Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes from the hbase API, or can I issue function call to get the progress of compaction?.
Hun

On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan <[EMAIL PROTECTED]> wrote:

> There is no interface which says that the major compaction is completed.
> But you can see that major compaction is in progress from the web UI.
> Sorry if am wrong here.
>
> Regards
> Ram
>
> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]> wrote:
>
>> Hi, All,
>> I want to measure the duration of a major compaction in HBase. Since the
>> function call majorCompact is asynchronous, I may need to manually check
>> when the major compaction is done. Does Hbase (as of version 0.92.4)
>> provide an interface to determine completion of major compaction? Thanks.
>> Yun
>>

Re: How to check if a major_compact is done?

Jean-Marc Spaggiari 2012-11-08, 11:43

Please someone correct me if I'm wrong, but I think there is some
information exposed to JMX which give you the duration (and size) of
the last compaction.

JM

2012/11/8, PG <[EMAIL PROTECTED]>:
> Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes from
> the hbase API, or can I issue function call to get the progress of
> compaction?.
> Hun
>
> On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
> <[EMAIL PROTECTED]> wrote:
>
>> There is no interface which says that the major compaction is completed.
>> But you can see that major compaction is in progress from the web UI.
>> Sorry if am wrong here.
>>
>> Regards
>> Ram
>>
>> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]> wrote:
>>
>>> Hi, All,
>>> I want to measure the duration of a major compaction in HBase. Since the
>>> function call majorCompact is asynchronous, I may need to manually check
>>> when the major compaction is done. Does Hbase (as of version 0.92.4)
>>> provide an interface to determine completion of major compaction?
>>> Thanks.
>>> Yun
>>>
>

Re: How to check if a major_compact is done?

Jerry Lam 2012-11-08, 15:43

Hi Yun:

Please refer to HBase Metric:
http://hbase.apache.org/book/hbase_metrics.html
The hbase.regionserver.compactionQueueSize seems promising but I'm not
certain because I have never use it.

Best Regards,

Jerry
On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <[EMAIL PROTECTED]
> wrote:

> Please someone correct me if I'm wrong, but I think there is some
> information exposed to JMX which give you the duration (and size) of
> the last compaction.
>
> JM
>
> 2012/11/8, PG <[EMAIL PROTECTED]>:
> > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes from
> > the hbase API, or can I issue function call to get the progress of
> > compaction?.
> > Hun
> >
> > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
> > <[EMAIL PROTECTED]> wrote:
> >
> >> There is no interface which says that the major compaction is completed.
> >> But you can see that major compaction is in progress from the web UI.
> >> Sorry if am wrong here.
> >>
> >> Regards
> >> Ram
> >>
> >> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]>
> wrote:
> >>
> >>> Hi, All,
> >>> I want to measure the duration of a major compaction in HBase. Since
> the
> >>> function call majorCompact is asynchronous, I may need to manually
> check
> >>> when the major compaction is done. Does Hbase (as of version 0.92.4)
> >>> provide an interface to determine completion of major compaction?
> >>> Thanks.
> >>> Yun
> >>>
> >
>

Re: How to check if a major_compact is done?

yun peng 2012-11-08, 18:31

Yes, JMX exposes compaction time. My cluster has JMX enabled, and I can
view certain statistics from remote jconsole.

However, I dont find any reference talking about how to write a Java
program to collect a specific statistics, like compaction time available in
region server status. Does anyone here have experience in writing low-level
code to collect JMX data, instead of using jconsole or other high-level
tools...

PS: I have found some sample code
here<http://hadoop.apache.org/docs/current/api/org/apache/hadoop/metrics/package-summary.html>,
yet the API is deprecated.
Regards,
Yun
On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <[EMAIL PROTECTED]
> wrote:

> Please someone correct me if I'm wrong, but I think there is some
> information exposed to JMX which give you the duration (and size) of
> the last compaction.
>
> JM
>
> 2012/11/8, PG <[EMAIL PROTECTED]>:
> > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes from
> > the hbase API, or can I issue function call to get the progress of
> > compaction?.
> > Hun
> >
> > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
> > <[EMAIL PROTECTED]> wrote:
> >
> >> There is no interface which says that the major compaction is completed.
> >> But you can see that major compaction is in progress from the web UI.
> >> Sorry if am wrong here.
> >>
> >> Regards
> >> Ram
> >>
> >> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]>
> wrote:
> >>
> >>> Hi, All,
> >>> I want to measure the duration of a major compaction in HBase. Since
> the
> >>> function call majorCompact is asynchronous, I may need to manually
> check
> >>> when the major compaction is done. Does Hbase (as of version 0.92.4)
> >>> provide an interface to determine completion of major compaction?
> >>> Thanks.
> >>> Yun
> >>>
> >
>

Re: How to check if a major_compact is done?

Otis Gospodnetic 2012-11-08, 22:03

Hi,

https://github.com/aravind/hbase_compact may have what you need (thanks
Jeremy)

Otis
--
Search Analytics - http://sematext.com/search-analytics/index.html
Performance Monitoring - http://sematext.com/spm/index.html
On Thu, Nov 8, 2012 at 1:31 PM, yun peng <[EMAIL PROTECTED]> wrote:

> Yes, JMX exposes compaction time. My cluster has JMX enabled, and I can
> view certain statistics from remote jconsole.
>
> However, I dont find any reference talking about how to write a Java
> program to collect a specific statistics, like compaction time available in
> region server status. Does anyone here have experience in writing low-level
> code to collect JMX data, instead of using jconsole or other high-level
> tools...
>
> PS: I have found some sample code
> here<
> http://hadoop.apache.org/docs/current/api/org/apache/hadoop/metrics/package-summary.html
> >,
> yet the API is deprecated.
> Regards,
> Yun
> On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <
> [EMAIL PROTECTED]
> > wrote:
>
> > Please someone correct me if I'm wrong, but I think there is some
> > information exposed to JMX which give you the duration (and size) of
> > the last compaction.
> >
> > JM
> >
> > 2012/11/8, PG <[EMAIL PROTECTED]>:
> > > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes
> from
> > > the hbase API, or can I issue function call to get the progress of
> > > compaction?.
> > > Hun
> > >
> > > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
> > > <[EMAIL PROTECTED]> wrote:
> > >
> > >> There is no interface which says that the major compaction is
> completed.
> > >> But you can see that major compaction is in progress from the web UI.
> > >> Sorry if am wrong here.
> > >>
> > >> Regards
> > >> Ram
> > >>
> > >> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]>
> > wrote:
> > >>
> > >>> Hi, All,
> > >>> I want to measure the duration of a major compaction in HBase. Since
> > the
> > >>> function call majorCompact is asynchronous, I may need to manually
> > check
> > >>> when the major compaction is done. Does Hbase (as of version 0.92.4)
> > >>> provide an interface to determine completion of major compaction?
> > >>> Thanks.
> > >>> Yun
> > >>>
> > >
> >
>

Re: How to check if a major_compact is done?

Jeremy Carroll 2012-11-08, 19:22

Sorry for spam. I have written metrics collection utilities via my GitHub
repo for JMX stats via the REST interface for each RegionServer.

https://github.com/phobos182/hadoop-hbase-tools

More specificially

https://github.com/phobos182/hadoop-hbase-tools/blob/master/hbase/regionserver_metrics.rb

With 0.94 changing the JMX stats to unify the web-ui stats, I will be
rewriting these to use JMX.

On Thu, Nov 8, 2012 at 10:31 AM, yun peng <[EMAIL PROTECTED]> wrote:

> Yes, JMX exposes compaction time. My cluster has JMX enabled, and I can
> view certain statistics from remote jconsole.
>
> However, I dont find any reference talking about how to write a Java
> program to collect a specific statistics, like compaction time available in
> region server status. Does anyone here have experience in writing low-level
> code to collect JMX data, instead of using jconsole or other high-level
> tools...
>
> PS: I have found some sample code
> here<
> http://hadoop.apache.org/docs/current/api/org/apache/hadoop/metrics/package-summary.html
> >,
> yet the API is deprecated.
> Regards,
> Yun
> On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <
> [EMAIL PROTECTED]
> > wrote:
>
> > Please someone correct me if I'm wrong, but I think there is some
> > information exposed to JMX which give you the duration (and size) of
> > the last compaction.
> >
> > JM
> >
> > 2012/11/8, PG <[EMAIL PROTECTED]>:
> > > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes
> from
> > > the hbase API, or can I issue function call to get the progress of
> > > compaction?.
> > > Hun
> > >
> > > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
> > > <[EMAIL PROTECTED]> wrote:
> > >
> > >> There is no interface which says that the major compaction is
> completed.
> > >> But you can see that major compaction is in progress from the web UI.
> > >> Sorry if am wrong here.
> > >>
> > >> Regards
> > >> Ram
> > >>
> > >> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]>
> > wrote:
> > >>
> > >>> Hi, All,
> > >>> I want to measure the duration of a major compaction in HBase. Since
> > the
> > >>> function call majorCompact is asynchronous, I may need to manually
> > check
> > >>> when the major compaction is done. Does Hbase (as of version 0.92.4)
> > >>> provide an interface to determine completion of major compaction?
> > >>> Thanks.
> > >>> Yun
> > >>>
> > >
> >
>

Re: How to check if a major_compact is done?

Jeremy Carroll 2012-11-08, 19:19

What I have done is using HBase 0.92.2+ (I believe commited in HBase 0.94)
there is a compaction status on the table via the Web-UI. I do not know if
it's exposed anywhere else. For my own internal compaction tools, I scrape
that status from the web UI to determine if a table is in minor / major /
none compaction.

https://s3.amazonaws.com/uploads.hipchat.com/10509/150181/39jhoklr70xbmfj/Screen%20Shot%202012-11-08%20at%2011.19.32%20AM.png
On Thu, Nov 8, 2012 at 10:31 AM, yun peng <[EMAIL PROTECTED]> wrote:

> Yes, JMX exposes compaction time. My cluster has JMX enabled, and I can
> view certain statistics from remote jconsole.
>
> However, I dont find any reference talking about how to write a Java
> program to collect a specific statistics, like compaction time available in
> region server status. Does anyone here have experience in writing low-level
> code to collect JMX data, instead of using jconsole or other high-level
> tools...
>
> PS: I have found some sample code
> here<
> http://hadoop.apache.org/docs/current/api/org/apache/hadoop/metrics/package-summary.html
> >,
> yet the API is deprecated.
> Regards,
> Yun
> On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <
> [EMAIL PROTECTED]
> > wrote:
>
> > Please someone correct me if I'm wrong, but I think there is some
> > information exposed to JMX which give you the duration (and size) of
> > the last compaction.
> >
> > JM
> >
> > 2012/11/8, PG <[EMAIL PROTECTED]>:
> > > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes
> from
> > > the hbase API, or can I issue function call to get the progress of
> > > compaction?.
> > > Hun
> > >
> > > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
> > > <[EMAIL PROTECTED]> wrote:
> > >
> > >> There is no interface which says that the major compaction is
> completed.
> > >> But you can see that major compaction is in progress from the web UI.
> > >> Sorry if am wrong here.
> > >>
> > >> Regards
> > >> Ram
> > >>
> > >> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]>
> > wrote:
> > >>
> > >>> Hi, All,
> > >>> I want to measure the duration of a major compaction in HBase. Since
> > the
> > >>> function call majorCompact is asynchronous, I may need to manually
> > check
> > >>> when the major compaction is done. Does Hbase (as of version 0.92.4)
> > >>> provide an interface to determine completion of major compaction?
> > >>> Thanks.
> > >>> Yun
> > >>>
> > >
> >
>

Re: How to check if a major_compact is done?

Jeremy Carroll 2012-11-08, 19:21

The JIRA ticket is as follows.

https://issues.apache.org/jira/browse/HBASE-6033 (Original)
https://issues.apache.org/jira/browse/HBASE-6124 (Backport to 0.92.1, etc..)

On Thu, Nov 8, 2012 at 11:19 AM, Jeremy Carroll <[EMAIL PROTECTED]> wrote:

> What I have done is using HBase 0.92.2+ (I believe commited in HBase 0.94)
> there is a compaction status on the table via the Web-UI. I do not know if
> it's exposed anywhere else. For my own internal compaction tools, I scrape
> that status from the web UI to determine if a table is in minor / major /
> none compaction.
>
>
> https://s3.amazonaws.com/uploads.hipchat.com/10509/150181/39jhoklr70xbmfj/Screen%20Shot%202012-11-08%20at%2011.19.32%20AM.png
>
>
> On Thu, Nov 8, 2012 at 10:31 AM, yun peng <[EMAIL PROTECTED]> wrote:
>
>> Yes, JMX exposes compaction time. My cluster has JMX enabled, and I can
>> view certain statistics from remote jconsole.
>>
>> However, I dont find any reference talking about how to write a Java
>> program to collect a specific statistics, like compaction time available
>> in
>> region server status. Does anyone here have experience in writing
>> low-level
>> code to collect JMX data, instead of using jconsole or other high-level
>> tools...
>>
>> PS: I have found some sample code
>> here<
>> http://hadoop.apache.org/docs/current/api/org/apache/hadoop/metrics/package-summary.html
>> >,
>> yet the API is deprecated.
>> Regards,
>> Yun
>> On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <
>> [EMAIL PROTECTED]
>> > wrote:
>>
>> > Please someone correct me if I'm wrong, but I think there is some
>> > information exposed to JMX which give you the duration (and size) of
>> > the last compaction.
>> >
>> > JM
>> >
>> > 2012/11/8, PG <[EMAIL PROTECTED]>:
>> > > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes
>> from
>> > > the hbase API, or can I issue function call to get the progress of
>> > > compaction?.
>> > > Hun
>> > >
>> > > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
>> > > <[EMAIL PROTECTED]> wrote:
>> > >
>> > >> There is no interface which says that the major compaction is
>> completed.
>> > >> But you can see that major compaction is in progress from the web UI.
>> > >> Sorry if am wrong here.
>> > >>
>> > >> Regards
>> > >> Ram
>> > >>
>> > >> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]>
>> > wrote:
>> > >>
>> > >>> Hi, All,
>> > >>> I want to measure the duration of a major compaction in HBase. Since
>> > the
>> > >>> function call majorCompact is asynchronous, I may need to manually
>> > check
>> > >>> when the major compaction is done. Does Hbase (as of version 0.92.4)
>> > >>> provide an interface to determine completion of major compaction?
>> > >>> Thanks.
>> > >>> Yun
>> > >>>
>> > >
>> >
>>
>
>

Re: How to check if a major_compact is done?

Mikael Sitruk 2012-11-08, 19:36

Yun hi

I did something like this in the past (in 0.92.0 version). That trigger
major compaction and monitor their completion. I did not go through the JMX
(which is not an indicator of major compaction), but via
the org.apache.hadoop.hbase.HServerLoad.RegionLoad
class.
When an major compaction ends the number of store file will be the number
of store which can be checked via: regionLoad.getStorefiles() =task.getStores()

I will try to find the code back and post it on github.
As for MBeans you can do somehting like:

url = new JMXServiceURL("service:jmx:rmi:///jndi/rmi://" + hostName + ":" +
portNum + "/jmxrmi");
HashMap<String, String> env = new HashMap<String, String>();
JMXConnector jmxc = JMXConnectorFactory.connect(url, env);
MBeanServerConnection mbsc = jmxc.getMBeanServerConnection();
ObjectName rsStat = new ObjectName(
"hadoop:service=RegionServer,name=RegionServerStatistics"); //put the
mbean you need here.
MBeanInfo mbeanInfo = mbsc.getMBeanInfo(rsStat);

for (MBeanAttributeInfo mbAttrInfo : mbeanInfo.getAttributes()) {
System.out.println(mbAttrInfo);
}
Regards,
Mikael.S

On Thu, Nov 8, 2012 at 8:31 PM, yun peng <[EMAIL PROTECTED]> wrote:

> Yes, JMX exposes compaction time. My cluster has JMX enabled, and I can
> view certain statistics from remote jconsole.
>
> However, I dont find any reference talking about how to write a Java
> program to collect a specific statistics, like compaction time available in
> region server status. Does anyone here have experience in writing low-level
> code to collect JMX data, instead of using jconsole or other high-level
> tools...
>
> PS: I have found some sample code
> here<
> http://hadoop.apache.org/docs/current/api/org/apache/hadoop/metrics/package-summary.html
> >,
> yet the API is deprecated.
> Regards,
> Yun
> On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <
> [EMAIL PROTECTED]
> > wrote:
>
> > Please someone correct me if I'm wrong, but I think there is some
> > information exposed to JMX which give you the duration (and size) of
> > the last compaction.
> >
> > JM
> >
> > 2012/11/8, PG <[EMAIL PROTECTED]>:
> > > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes
> from
> > > the hbase API, or can I issue function call to get the progress of
> > > compaction?.
> > > Hun
> > >
> > > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
> > > <[EMAIL PROTECTED]> wrote:
> > >
> > >> There is no interface which says that the major compaction is
> completed.
> > >> But you can see that major compaction is in progress from the web UI.
> > >> Sorry if am wrong here.
> > >>
> > >> Regards
> > >> Ram
> > >>
> > >> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]>
> > wrote:
> > >>
> > >>> Hi, All,
> > >>> I want to measure the duration of a major compaction in HBase. Since
> > the
> > >>> function call majorCompact is asynchronous, I may need to manually
> > check
> > >>> when the major compaction is done. Does Hbase (as of version 0.92.4)
> > >>> provide an interface to determine completion of major compaction?
> > >>> Thanks.
> > >>> Yun
> > >>>
> > >
> >
>

Re: How to check if a major_compact is done?

yun peng 2012-11-08, 20:58

Hi, Mikael and Jeremy, thanks for your detailed answers. I have tried the
MBean code as an attempt. I get certain information out from the printout,
like follows,
...
javax.management.MBeanAttributeInfo[description=NoDescription,
name=compactionTimeNumOps, type=java.lang.Integer, read-only, descriptor={}]
...

I have a minor question, (sorry about the extra questions, but I am sort of
newbie to web service interfaces). The descriptor (or the content) of
MBeanAttributeInfo here is empty, which does not match with the data I can
see through jconsole, which show compactionTimeNumOps=4 (after issuing two
major compaction requests). Should I pause the data/content from somewhere
other than MBeanAttributeInfo.getDescriptor().

Regards,
Yun
On Thu, Nov 8, 2012 at 2:36 PM, Mikael Sitruk <[EMAIL PROTECTED]>wrote:

> Yun hi
>
> I did something like this in the past (in 0.92.0 version). That trigger
> major compaction and monitor their completion. I did not go through the JMX
> (which is not an indicator of major compaction), but via
> the org.apache.hadoop.hbase.HServerLoad.RegionLoad
> class.
> When an major compaction ends the number of store file will be the number
> of store which can be checked via: regionLoad.getStorefiles() => task.getStores()
>
> I will try to find the code back and post it on github.
>
>
> As for MBeans you can do somehting like:
>
> url = new JMXServiceURL("service:jmx:rmi:///jndi/rmi://" + hostName + ":" +
> portNum + "/jmxrmi");
> HashMap<String, String> env = new HashMap<String, String>();
> JMXConnector jmxc = JMXConnectorFactory.connect(url, env);
> MBeanServerConnection mbsc = jmxc.getMBeanServerConnection();
> ObjectName rsStat = new ObjectName(
> "hadoop:service=RegionServer,name=RegionServerStatistics"); //put the
> mbean you need here.
> MBeanInfo mbeanInfo = mbsc.getMBeanInfo(rsStat);
>
> for (MBeanAttributeInfo mbAttrInfo : mbeanInfo.getAttributes()) {
> System.out.println(mbAttrInfo);
> }
>
>
> Regards,
> Mikael.S
>
> On Thu, Nov 8, 2012 at 8:31 PM, yun peng <[EMAIL PROTECTED]> wrote:
>
> > Yes, JMX exposes compaction time. My cluster has JMX enabled, and I can
> > view certain statistics from remote jconsole.
> >
> > However, I dont find any reference talking about how to write a Java
> > program to collect a specific statistics, like compaction time available
> in
> > region server status. Does anyone here have experience in writing
> low-level
> > code to collect JMX data, instead of using jconsole or other high-level
> > tools...
> >
> > PS: I have found some sample code
> > here<
> >
> http://hadoop.apache.org/docs/current/api/org/apache/hadoop/metrics/package-summary.html
> > >,
> > yet the API is deprecated.
> > Regards,
> > Yun
> > On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <
> > [EMAIL PROTECTED]
> > > wrote:
> >
> > > Please someone correct me if I'm wrong, but I think there is some
> > > information exposed to JMX which give you the duration (and size) of
> > > the last compaction.
> > >
> > > JM
> > >
> > > 2012/11/8, PG <[EMAIL PROTECTED]>:
> > > > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes
> > from
> > > > the hbase API, or can I issue function call to get the progress of
> > > > compaction?.
> > > > Hun
> > > >
> > > > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
> > > > <[EMAIL PROTECTED]> wrote:
> > > >
> > > >> There is no interface which says that the major compaction is
> > completed.
> > > >> But you can see that major compaction is in progress from the web
> UI.
> > > >> Sorry if am wrong here.
> > > >>
> > > >> Regards
> > > >> Ram
> > > >>
> > > >> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]>
> > > wrote:
> > > >>
> > > >>> Hi, All,
> > > >>> I want to measure the duration of a major compaction in HBase.
> Since
> > > the
> > > >>> function call majorCompact is asynchronous, I may need to manually
> > > check
> > > >>> when the major compaction is done. Does Hbase (as of version

Re: How to check if a major_compact is done?

yun peng 2012-11-08, 21:08

Hi, I have fixed the problem. I found the
MBeanServerConnection.getAttribute(ObjectName name, String attribute) is
the one useful here. Thanks to all.
Yun

On Thu, Nov 8, 2012 at 3:58 PM, yun peng <[EMAIL PROTECTED]> wrote:

> Hi, Mikael and Jeremy, thanks for your detailed answers. I have tried the
> MBean code as an attempt. I get certain information out from the printout,
> like follows,
> ...
> javax.management.MBeanAttributeInfo[description=NoDescription,
> name=compactionTimeNumOps, type=java.lang.Integer, read-only, descriptor={}]
> ...
>
> I have a minor question, (sorry about the extra questions, but I am sort
> of newbie to web service interfaces). The descriptor (or the content) of
> MBeanAttributeInfo here is empty, which does not match with the data I can
> see through jconsole, which show compactionTimeNumOps=4 (after issuing two
> major compaction requests). Should I pause the data/content from somewhere
> other than MBeanAttributeInfo.getDescriptor().
>
> Regards,
> Yun
>
>
>
> On Thu, Nov 8, 2012 at 2:36 PM, Mikael Sitruk <[EMAIL PROTECTED]>wrote:
>
>> Yun hi
>>
>> I did something like this in the past (in 0.92.0 version). That trigger
>> major compaction and monitor their completion. I did not go through the
>> JMX
>> (which is not an indicator of major compaction), but via
>> the org.apache.hadoop.hbase.HServerLoad.RegionLoad
>> class.
>> When an major compaction ends the number of store file will be the number
>> of store which can be checked via: regionLoad.getStorefiles() =>> task.getStores()
>>
>> I will try to find the code back and post it on github.
>>
>>
>> As for MBeans you can do somehting like:
>>
>> url = new JMXServiceURL("service:jmx:rmi:///jndi/rmi://" + hostName + ":"
>> +
>> portNum + "/jmxrmi");
>> HashMap<String, String> env = new HashMap<String, String>();
>> JMXConnector jmxc = JMXConnectorFactory.connect(url, env);
>> MBeanServerConnection mbsc = jmxc.getMBeanServerConnection();
>> ObjectName rsStat = new ObjectName(
>> "hadoop:service=RegionServer,name=RegionServerStatistics"); //put the
>> mbean you need here.
>> MBeanInfo mbeanInfo = mbsc.getMBeanInfo(rsStat);
>>
>> for (MBeanAttributeInfo mbAttrInfo : mbeanInfo.getAttributes()) {
>> System.out.println(mbAttrInfo);
>> }
>>
>>
>> Regards,
>> Mikael.S
>>
>> On Thu, Nov 8, 2012 at 8:31 PM, yun peng <[EMAIL PROTECTED]> wrote:
>>
>> > Yes, JMX exposes compaction time. My cluster has JMX enabled, and I can
>> > view certain statistics from remote jconsole.
>> >
>> > However, I dont find any reference talking about how to write a Java
>> > program to collect a specific statistics, like compaction time
>> available in
>> > region server status. Does anyone here have experience in writing
>> low-level
>> > code to collect JMX data, instead of using jconsole or other high-level
>> > tools...
>> >
>> > PS: I have found some sample code
>> > here<
>> >
>> http://hadoop.apache.org/docs/current/api/org/apache/hadoop/metrics/package-summary.html
>> > >,
>> > yet the API is deprecated.
>> > Regards,
>> > Yun
>> > On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <
>> > [EMAIL PROTECTED]
>> > > wrote:
>> >
>> > > Please someone correct me if I'm wrong, but I think there is some
>> > > information exposed to JMX which give you the duration (and size) of
>> > > the last compaction.
>> > >
>> > > JM
>> > >
>> > > 2012/11/8, PG <[EMAIL PROTECTED]>:
>> > > > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes
>> > from
>> > > > the hbase API, or can I issue function call to get the progress of
>> > > > compaction?.
>> > > > Hun
>> > > >
>> > > > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
>> > > > <[EMAIL PROTECTED]> wrote:
>> > > >
>> > > >> There is no interface which says that the major compaction is
>> > completed.
>> > > >> But you can see that major compaction is in progress from the web
>> UI.
>> > > >> Sorry if am wrong here.
>> > > >>
>> > > >> Regards
>> > > >> Ram
>> > > >>
>> >

Re: How to check if a major_compact is done?

yun peng 2012-11-17, 01:03

Hi, as I recently find out, by spinning on getStorefiles =task.getStores() may not be a robust method to check if
a compaction is done. For instance, if the interval for checking is set too
large and there are compaction and flushes
intertwined together, then in the end, getStorefiles() may not be equal to
task.getStores() even when compaction is done.

Currently, I have to disable flush(), but am open for other solutions....
Regards,
Yun

On Thu, Nov 8, 2012 at 2:36 PM, Mikael Sitruk <[EMAIL PROTECTED]>wrote:

> Yun hi
>
> I did something like this in the past (in 0.92.0 version). That trigger
> major compaction and monitor their completion. I did not go through the JMX
> (which is not an indicator of major compaction), but via
> the org.apache.hadoop.hbase.HServerLoad.RegionLoad
> class.
> When an major compaction ends the number of store file will be the number
> of store which can be checked via: regionLoad.getStorefiles() => task.getStores()
>
> I will try to find the code back and post it on github.
>
>
> As for MBeans you can do somehting like:
>
> url = new JMXServiceURL("service:jmx:rmi:///jndi/rmi://" + hostName + ":" +
> portNum + "/jmxrmi");
> HashMap<String, String> env = new HashMap<String, String>();
> JMXConnector jmxc = JMXConnectorFactory.connect(url, env);
> MBeanServerConnection mbsc = jmxc.getMBeanServerConnection();
> ObjectName rsStat = new ObjectName(
> "hadoop:service=RegionServer,name=RegionServerStatistics"); //put the
> mbean you need here.
> MBeanInfo mbeanInfo = mbsc.getMBeanInfo(rsStat);
>
> for (MBeanAttributeInfo mbAttrInfo : mbeanInfo.getAttributes()) {
> System.out.println(mbAttrInfo);
> }
>
>
> Regards,
> Mikael.S
>
> On Thu, Nov 8, 2012 at 8:31 PM, yun peng <[EMAIL PROTECTED]> wrote:
>
> > Yes, JMX exposes compaction time. My cluster has JMX enabled, and I can
> > view certain statistics from remote jconsole.
> >
> > However, I dont find any reference talking about how to write a Java
> > program to collect a specific statistics, like compaction time available
> in
> > region server status. Does anyone here have experience in writing
> low-level
> > code to collect JMX data, instead of using jconsole or other high-level
> > tools...
> >
> > PS: I have found some sample code
> > here<
> >
> http://hadoop.apache.org/docs/current/api/org/apache/hadoop/metrics/package-summary.html
> > >,
> > yet the API is deprecated.
> > Regards,
> > Yun
> > On Thu, Nov 8, 2012 at 6:43 AM, Jean-Marc Spaggiari <
> > [EMAIL PROTECTED]
> > > wrote:
> >
> > > Please someone correct me if I'm wrong, but I think there is some
> > > information exposed to JMX which give you the duration (and size) of
> > > the last compaction.
> > >
> > > JM
> > >
> > > 2012/11/8, PG <[EMAIL PROTECTED]>:
> > > > Hi, thanks for the comments. One thing is,,,, shouldn't web UI comes
> > from
> > > > the hbase API, or can I issue function call to get the progress of
> > > > compaction?.
> > > > Hun
> > > >
> > > > On Nov 8, 2012, at 1:33 AM, ramkrishna vasudevan
> > > > <[EMAIL PROTECTED]> wrote:
> > > >
> > > >> There is no interface which says that the major compaction is
> > completed.
> > > >> But you can see that major compaction is in progress from the web
> UI.
> > > >> Sorry if am wrong here.
> > > >>
> > > >> Regards
> > > >> Ram
> > > >>
> > > >> On Thu, Nov 8, 2012 at 11:38 AM, yun peng <[EMAIL PROTECTED]>
> > > wrote:
> > > >>
> > > >>> Hi, All,
> > > >>> I want to measure the duration of a major compaction in HBase.
> Since
> > > the
> > > >>> function call majorCompact is asynchronous, I may need to manually
> > > check
> > > >>> when the major compaction is done. Does Hbase (as of version
> 0.92.4)
> > > >>> provide an interface to determine completion of major compaction?
> > > >>> Thanks.
> > > >>> Yun
> > > >>>
> > > >
> > >
> >
>

Re: How to check if a major_compact is done?

Otis Gospodnetic 2012-11-08, 16:27

Hi,

I *think* SPM for HBase shows this. URL in signature. I'm guessing
Cloudera Manager may also show this if you are using CDH. Otherwise, you
can just read from JMX. Handy no-deps command-line tool for that:
https://github.com/sematext/jmxc

Otis
--
Performance Monitoring - http://sematext.com/spm/index.html
On Thu, Nov 8, 2012 at 1:08 AM, yun peng <[EMAIL PROTECTED]> wrote:

> Hi, All,
> I want to measure the duration of a major compaction in HBase. Since the
> function call majorCompact is asynchronous, I may need to manually check
> when the major compaction is done. Does Hbase (as of version 0.92.4)
> provide an interface to determine completion of major compaction? Thanks.
> Yun
>