You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hbase.apache.org by Weihua JIANG <we...@gmail.com> on 2012/01/05 07:06:08 UTC
Is hadoop 1.0.0 + HBase 0.90.5 the best combination for production cluster?
Hi all,
Hadoop 1.0.0 and HBase 0.90.5 are released. I am curious whether hadoop
1.0.0 is compatible with HBase 0.90.5. And whether this combination is the
best choice for production cluster now?
Thanks
Weihua
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for
production cluster?
Posted by Jean-Daniel Cryans <jd...@apache.org>.
The gist of the answer is that, unlike random reads, the blocks we
read sequentially from the fs are wholly consumed so you end up doing
less fs calls thus the total proportion of the time spent talking to
datanodes is lessened (which is what local reads help).
Also the dfs client keeps a block reader opened so that every time you
read from the same hdfs block it doesn't have to setup the socket to
the datanode again (which is what random reading does if you don't
setup local reads).
J-D
On Fri, Feb 17, 2012 at 1:48 PM, Bryan Keller <br...@gmail.com> wrote:
> I was thinking (wrongly it seems) that having the region server read directly from the local file system would be faster than going through the data node, even with sequential access.
>
> On Feb 17, 2012, at 1:28 PM, Jean-Daniel Cryans wrote:
>
>> On Fri, Feb 17, 2012 at 1:21 PM, Bryan Keller <br...@gmail.com> wrote:
>>> I have been experimenting with local reads. For me, enabling did not help improve read performance at all, I get the same performance either way. I can see in the data node logs it is passing back the local path, so it is enabled properly.
>>
>> I was surprised when I read this until I saw this:
>>
>>>
>>> Perhaps the benefits of local reads are dependent on the type of data and the workload? In my test I'm scanning through the entire table via a map reduce job. It's a wide table with maybe 20k columns per row on average. I have scanner caching set to 10.
>>
>> It's definitely not going to help make sequential reads faster.
>>
>>>
>>> My read performance is about 10% of the disk max read throughput, i.e. my disks can get 100 mb/sec tested with hdparm and scan performance is about 10 mb/sec. Not too bad I suppose.
>>
>> Maybe you're not pushing it enough?
>>
>> J-D
>
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for production cluster?
Posted by Bryan Keller <br...@gmail.com>.
I was thinking (wrongly it seems) that having the region server read directly from the local file system would be faster than going through the data node, even with sequential access.
On Feb 17, 2012, at 1:28 PM, Jean-Daniel Cryans wrote:
> On Fri, Feb 17, 2012 at 1:21 PM, Bryan Keller <br...@gmail.com> wrote:
>> I have been experimenting with local reads. For me, enabling did not help improve read performance at all, I get the same performance either way. I can see in the data node logs it is passing back the local path, so it is enabled properly.
>
> I was surprised when I read this until I saw this:
>
>>
>> Perhaps the benefits of local reads are dependent on the type of data and the workload? In my test I'm scanning through the entire table via a map reduce job. It's a wide table with maybe 20k columns per row on average. I have scanner caching set to 10.
>
> It's definitely not going to help make sequential reads faster.
>
>>
>> My read performance is about 10% of the disk max read throughput, i.e. my disks can get 100 mb/sec tested with hdparm and scan performance is about 10 mb/sec. Not too bad I suppose.
>
> Maybe you're not pushing it enough?
>
> J-D
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for
production cluster?
Posted by Jean-Daniel Cryans <jd...@apache.org>.
On Fri, Feb 17, 2012 at 1:21 PM, Bryan Keller <br...@gmail.com> wrote:
> I have been experimenting with local reads. For me, enabling did not help improve read performance at all, I get the same performance either way. I can see in the data node logs it is passing back the local path, so it is enabled properly.
I was surprised when I read this until I saw this:
>
> Perhaps the benefits of local reads are dependent on the type of data and the workload? In my test I'm scanning through the entire table via a map reduce job. It's a wide table with maybe 20k columns per row on average. I have scanner caching set to 10.
It's definitely not going to help make sequential reads faster.
>
> My read performance is about 10% of the disk max read throughput, i.e. my disks can get 100 mb/sec tested with hdparm and scan performance is about 10 mb/sec. Not too bad I suppose.
Maybe you're not pushing it enough?
J-D
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for production cluster?
Posted by Bryan Keller <br...@gmail.com>.
I have been experimenting with local reads. For me, enabling did not help improve read performance at all, I get the same performance either way. I can see in the data node logs it is passing back the local path, so it is enabled properly.
Perhaps the benefits of local reads are dependent on the type of data and the workload? In my test I'm scanning through the entire table via a map reduce job. It's a wide table with maybe 20k columns per row on average. I have scanner caching set to 10.
My read performance is about 10% of the disk max read throughput, i.e. my disks can get 100 mb/sec tested with hdparm and scan performance is about 10 mb/sec. Not too bad I suppose.
On Jan 8, 2012, at 6:35 PM, Zizon Qiu wrote:
> It should be the same as hbase daemon user.
>
> the check perform by datanode are implement as follow, inside a RPC call.
> the "current user" refer to the remote user,in this case, should the same
> as your hbase user
>
> private void checkBlockLocalPathAccess() throws IOException {
> checkKerberosAuthMethod("getBlockLocalPathInfo()");
> *String currentUser =
> UserGroupInformation.getCurrentUser().getShortUserName();*
> if (!*currentUser*.equals(this.userWithLocalPathAccess)) {
> throw new AccessControlException(
> "Can't continue with getBlockLocalPathInfo() "
> + "authorization. The user " + currentUser
> + " is not allowed to call getBlockLocalPathInfo");
> }
> }
>
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for
production cluster?
Posted by Zizon Qiu <zz...@gmail.com>.
It should be the same as hbase daemon user.
the check perform by datanode are implement as follow, inside a RPC call.
the "current user" refer to the remote user,in this case, should the same
as your hbase user
private void checkBlockLocalPathAccess() throws IOException {
checkKerberosAuthMethod("getBlockLocalPathInfo()");
*String currentUser =
UserGroupInformation.getCurrentUser().getShortUserName();*
if (!*currentUser*.equals(this.userWithLocalPathAccess)) {
throw new AccessControlException(
"Can't continue with getBlockLocalPathInfo() "
+ "authorization. The user " + currentUser
+ " is not allowed to call getBlockLocalPathInfo");
}
}
On Sun, Jan 8, 2012 at 11:45 PM, Yves Langisch <yv...@langisch.ch> wrote:
> I have no special settings for the hadoop security. Is it necessary to
> specify dfs.block.local-path-access.user then? If yes, is it the same user
> my hadoop daemon is running under?
>
> Thanks
> Yves
>
> On Jan 6, 2012, at 4:27 PM, Stack wrote:
>
> > On Thu, Jan 5, 2012 at 10:48 PM, zizon <zz...@gmail.com> wrote:
> >> <property>
> >> <name>*dfs.client.read.shortcircuit*</name>
> >> <value>true</value>
> >> <description>set this to true to enable DFSClient short circuit
> >> read</description>
> >> </property>
> >>
> >
> > You must set the above on both client and server side; i.e. in
> > hbase-site.xml and in hdfs-site.xml.
> >
> >> <property>
> >> <name>*dfs.block.local-path-access.user*</name>
> >> <value>hadoop</value>
> >> <description>add users that need perform short circuit read
> here,datanode
> >> will do security check before the read</description>
> >> </property>
> >>
> >
> > This you set server-side only. Adjust the value 'hadoop' accordingly.
> >
> > I believe the only way to tell local reads are working is jstacking
> > your hbase and looking for local block accesses.
> >
> > We''ll fix our documentation to include the above.
> >
> > St.Ack
> >
> >>
> >> On Fri, Jan 6, 2012 at 2:25 PM, Yves Langisch <yv...@langisch.ch> wrote:
> >>
> >>> How can you enable the mentioned local-read-optimization for
> hadoop-1.0.0?
> >>> I could not find any related information.
> >>>
> >>> -
> >>> Yves
> >>>
> >>> On Jan 6, 2012, at 6:24 AM, Arun C Murthy wrote:
> >>>
> >>>> I know we've done integration testing with hadoop-1.0.0 and
> hbase-0.90.4
> >>> and things work well, not sure about hbase-0.90.5 (I don't imagine
> there
> >>> are issues, but ymmv).
> >>>>
> >>>> In fact, you get a nice perf boost with hadoop-1.0.0 for hbase if you
> >>> enable the local-read-optimization.
> >>>>
> >>>> Arun
> >>>>
> >>>> On Jan 4, 2012, at 10:19 PM, praveenesh kumar wrote:
> >>>>
> >>>>> Don't know about Hadoop 1.0.0 but Hadoop 0.20.205 and Hadoop 0.90.5
> are
> >>>>> playing with each other very fine.
> >>>>>
> >>>>> Thanks,
> >>>>> Praveenesh
> >>>>>
> >>>>> On Thu, Jan 5, 2012 at 11:36 AM, Weihua JIANG <
> weihua.jiang@gmail.com
> >>>> wrote:
> >>>>>
> >>>>>> Hi all,
> >>>>>>
> >>>>>> Hadoop 1.0.0 and HBase 0.90.5 are released. I am curious whether
> hadoop
> >>>>>> 1.0.0 is compatible with HBase 0.90.5. And whether this combination
> is
> >>> the
> >>>>>> best choice for production cluster now?
> >>>>>>
> >>>>>> Thanks
> >>>>>> Weihua
> >>>>>>
> >>>>
> >>>
> >>>
> >
>
>
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for production cluster?
Posted by Yves Langisch <yv...@langisch.ch>.
I have no special settings for the hadoop security. Is it necessary to specify dfs.block.local-path-access.user then? If yes, is it the same user my hadoop daemon is running under?
Thanks
Yves
On Jan 6, 2012, at 4:27 PM, Stack wrote:
> On Thu, Jan 5, 2012 at 10:48 PM, zizon <zz...@gmail.com> wrote:
>> <property>
>> <name>*dfs.client.read.shortcircuit*</name>
>> <value>true</value>
>> <description>set this to true to enable DFSClient short circuit
>> read</description>
>> </property>
>>
>
> You must set the above on both client and server side; i.e. in
> hbase-site.xml and in hdfs-site.xml.
>
>> <property>
>> <name>*dfs.block.local-path-access.user*</name>
>> <value>hadoop</value>
>> <description>add users that need perform short circuit read here,datanode
>> will do security check before the read</description>
>> </property>
>>
>
> This you set server-side only. Adjust the value 'hadoop' accordingly.
>
> I believe the only way to tell local reads are working is jstacking
> your hbase and looking for local block accesses.
>
> We''ll fix our documentation to include the above.
>
> St.Ack
>
>>
>> On Fri, Jan 6, 2012 at 2:25 PM, Yves Langisch <yv...@langisch.ch> wrote:
>>
>>> How can you enable the mentioned local-read-optimization for hadoop-1.0.0?
>>> I could not find any related information.
>>>
>>> -
>>> Yves
>>>
>>> On Jan 6, 2012, at 6:24 AM, Arun C Murthy wrote:
>>>
>>>> I know we've done integration testing with hadoop-1.0.0 and hbase-0.90.4
>>> and things work well, not sure about hbase-0.90.5 (I don't imagine there
>>> are issues, but ymmv).
>>>>
>>>> In fact, you get a nice perf boost with hadoop-1.0.0 for hbase if you
>>> enable the local-read-optimization.
>>>>
>>>> Arun
>>>>
>>>> On Jan 4, 2012, at 10:19 PM, praveenesh kumar wrote:
>>>>
>>>>> Don't know about Hadoop 1.0.0 but Hadoop 0.20.205 and Hadoop 0.90.5 are
>>>>> playing with each other very fine.
>>>>>
>>>>> Thanks,
>>>>> Praveenesh
>>>>>
>>>>> On Thu, Jan 5, 2012 at 11:36 AM, Weihua JIANG <weihua.jiang@gmail.com
>>>> wrote:
>>>>>
>>>>>> Hi all,
>>>>>>
>>>>>> Hadoop 1.0.0 and HBase 0.90.5 are released. I am curious whether hadoop
>>>>>> 1.0.0 is compatible with HBase 0.90.5. And whether this combination is
>>> the
>>>>>> best choice for production cluster now?
>>>>>>
>>>>>> Thanks
>>>>>> Weihua
>>>>>>
>>>>
>>>
>>>
>
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for
production cluster?
Posted by Stack <st...@duboce.net>.
On Thu, Jan 5, 2012 at 10:48 PM, zizon <zz...@gmail.com> wrote:
> <property>
> <name>*dfs.client.read.shortcircuit*</name>
> <value>true</value>
> <description>set this to true to enable DFSClient short circuit
> read</description>
> </property>
>
You must set the above on both client and server side; i.e. in
hbase-site.xml and in hdfs-site.xml.
> <property>
> <name>*dfs.block.local-path-access.user*</name>
> <value>hadoop</value>
> <description>add users that need perform short circuit read here,datanode
> will do security check before the read</description>
> </property>
>
This you set server-side only. Adjust the value 'hadoop' accordingly.
I believe the only way to tell local reads are working is jstacking
your hbase and looking for local block accesses.
We''ll fix our documentation to include the above.
St.Ack
>
> On Fri, Jan 6, 2012 at 2:25 PM, Yves Langisch <yv...@langisch.ch> wrote:
>
>> How can you enable the mentioned local-read-optimization for hadoop-1.0.0?
>> I could not find any related information.
>>
>> -
>> Yves
>>
>> On Jan 6, 2012, at 6:24 AM, Arun C Murthy wrote:
>>
>> > I know we've done integration testing with hadoop-1.0.0 and hbase-0.90.4
>> and things work well, not sure about hbase-0.90.5 (I don't imagine there
>> are issues, but ymmv).
>> >
>> > In fact, you get a nice perf boost with hadoop-1.0.0 for hbase if you
>> enable the local-read-optimization.
>> >
>> > Arun
>> >
>> > On Jan 4, 2012, at 10:19 PM, praveenesh kumar wrote:
>> >
>> >> Don't know about Hadoop 1.0.0 but Hadoop 0.20.205 and Hadoop 0.90.5 are
>> >> playing with each other very fine.
>> >>
>> >> Thanks,
>> >> Praveenesh
>> >>
>> >> On Thu, Jan 5, 2012 at 11:36 AM, Weihua JIANG <weihua.jiang@gmail.com
>> >wrote:
>> >>
>> >>> Hi all,
>> >>>
>> >>> Hadoop 1.0.0 and HBase 0.90.5 are released. I am curious whether hadoop
>> >>> 1.0.0 is compatible with HBase 0.90.5. And whether this combination is
>> the
>> >>> best choice for production cluster now?
>> >>>
>> >>> Thanks
>> >>> Weihua
>> >>>
>> >
>>
>>
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for
production cluster?
Posted by zizon <zz...@gmail.com>.
<property>
<name>*dfs.client.read.shortcircuit*</name>
<value>true</value>
<description>set this to true to enable DFSClient short circuit
read</description>
</property>
<property>
<name>*dfs.block.local-path-access.user*</name>
<value>hadoop</value>
<description>add users that need perform short circuit read here,datanode
will do security check before the read</description>
</property>
On Fri, Jan 6, 2012 at 2:25 PM, Yves Langisch <yv...@langisch.ch> wrote:
> How can you enable the mentioned local-read-optimization for hadoop-1.0.0?
> I could not find any related information.
>
> -
> Yves
>
> On Jan 6, 2012, at 6:24 AM, Arun C Murthy wrote:
>
> > I know we've done integration testing with hadoop-1.0.0 and hbase-0.90.4
> and things work well, not sure about hbase-0.90.5 (I don't imagine there
> are issues, but ymmv).
> >
> > In fact, you get a nice perf boost with hadoop-1.0.0 for hbase if you
> enable the local-read-optimization.
> >
> > Arun
> >
> > On Jan 4, 2012, at 10:19 PM, praveenesh kumar wrote:
> >
> >> Don't know about Hadoop 1.0.0 but Hadoop 0.20.205 and Hadoop 0.90.5 are
> >> playing with each other very fine.
> >>
> >> Thanks,
> >> Praveenesh
> >>
> >> On Thu, Jan 5, 2012 at 11:36 AM, Weihua JIANG <weihua.jiang@gmail.com
> >wrote:
> >>
> >>> Hi all,
> >>>
> >>> Hadoop 1.0.0 and HBase 0.90.5 are released. I am curious whether hadoop
> >>> 1.0.0 is compatible with HBase 0.90.5. And whether this combination is
> the
> >>> best choice for production cluster now?
> >>>
> >>> Thanks
> >>> Weihua
> >>>
> >
>
>
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for production cluster?
Posted by Harsh J <ha...@cloudera.com>.
Yves,
Take a look at the release notes section of https://issues.apache.org/jira/browse/HDFS-2246 for the configs that are relevant to that feature addition.
On 06-Jan-2012, at 11:55 AM, Yves Langisch wrote:
> How can you enable the mentioned local-read-optimization for hadoop-1.0.0? I could not find any related information.
>
> -
> Yves
>
> On Jan 6, 2012, at 6:24 AM, Arun C Murthy wrote:
>
>> I know we've done integration testing with hadoop-1.0.0 and hbase-0.90.4 and things work well, not sure about hbase-0.90.5 (I don't imagine there are issues, but ymmv).
>>
>> In fact, you get a nice perf boost with hadoop-1.0.0 for hbase if you enable the local-read-optimization.
>>
>> Arun
>>
>> On Jan 4, 2012, at 10:19 PM, praveenesh kumar wrote:
>>
>>> Don't know about Hadoop 1.0.0 but Hadoop 0.20.205 and Hadoop 0.90.5 are
>>> playing with each other very fine.
>>>
>>> Thanks,
>>> Praveenesh
>>>
>>> On Thu, Jan 5, 2012 at 11:36 AM, Weihua JIANG <we...@gmail.com>wrote:
>>>
>>>> Hi all,
>>>>
>>>> Hadoop 1.0.0 and HBase 0.90.5 are released. I am curious whether hadoop
>>>> 1.0.0 is compatible with HBase 0.90.5. And whether this combination is the
>>>> best choice for production cluster now?
>>>>
>>>> Thanks
>>>> Weihua
>>>>
>>
>
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for production cluster?
Posted by Yves Langisch <yv...@langisch.ch>.
How can you enable the mentioned local-read-optimization for hadoop-1.0.0? I could not find any related information.
-
Yves
On Jan 6, 2012, at 6:24 AM, Arun C Murthy wrote:
> I know we've done integration testing with hadoop-1.0.0 and hbase-0.90.4 and things work well, not sure about hbase-0.90.5 (I don't imagine there are issues, but ymmv).
>
> In fact, you get a nice perf boost with hadoop-1.0.0 for hbase if you enable the local-read-optimization.
>
> Arun
>
> On Jan 4, 2012, at 10:19 PM, praveenesh kumar wrote:
>
>> Don't know about Hadoop 1.0.0 but Hadoop 0.20.205 and Hadoop 0.90.5 are
>> playing with each other very fine.
>>
>> Thanks,
>> Praveenesh
>>
>> On Thu, Jan 5, 2012 at 11:36 AM, Weihua JIANG <we...@gmail.com>wrote:
>>
>>> Hi all,
>>>
>>> Hadoop 1.0.0 and HBase 0.90.5 are released. I am curious whether hadoop
>>> 1.0.0 is compatible with HBase 0.90.5. And whether this combination is the
>>> best choice for production cluster now?
>>>
>>> Thanks
>>> Weihua
>>>
>
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for production cluster?
Posted by Arun C Murthy <ac...@hortonworks.com>.
I know we've done integration testing with hadoop-1.0.0 and hbase-0.90.4 and things work well, not sure about hbase-0.90.5 (I don't imagine there are issues, but ymmv).
In fact, you get a nice perf boost with hadoop-1.0.0 for hbase if you enable the local-read-optimization.
Arun
On Jan 4, 2012, at 10:19 PM, praveenesh kumar wrote:
> Don't know about Hadoop 1.0.0 but Hadoop 0.20.205 and Hadoop 0.90.5 are
> playing with each other very fine.
>
> Thanks,
> Praveenesh
>
> On Thu, Jan 5, 2012 at 11:36 AM, Weihua JIANG <we...@gmail.com>wrote:
>
>> Hi all,
>>
>> Hadoop 1.0.0 and HBase 0.90.5 are released. I am curious whether hadoop
>> 1.0.0 is compatible with HBase 0.90.5. And whether this combination is the
>> best choice for production cluster now?
>>
>> Thanks
>> Weihua
>>
Re: Is hadoop 1.0.0 + HBase 0.90.5 the best combination for
production cluster?
Posted by praveenesh kumar <pr...@gmail.com>.
Don't know about Hadoop 1.0.0 but Hadoop 0.20.205 and Hadoop 0.90.5 are
playing with each other very fine.
Thanks,
Praveenesh
On Thu, Jan 5, 2012 at 11:36 AM, Weihua JIANG <we...@gmail.com>wrote:
> Hi all,
>
> Hadoop 1.0.0 and HBase 0.90.5 are released. I am curious whether hadoop
> 1.0.0 is compatible with HBase 0.90.5. And whether this combination is the
> best choice for production cluster now?
>
> Thanks
> Weihua
>