You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hama.apache.org by Júlio Pires <ju...@gmail.com> on 2013/09/03 20:58:26 UTC

Question About Distributed Mode

Hello,

I have three machines, named as follows: master, slave_a, slave_b

Questions:
1) In the files "groomservers" and hama-site.xml (property
hama.zookeeper.quorum ), should be included the three machines or just
slave_a and slave_b?

2) With the running environment, how do I add a new slave machine
On-the-fly?

Thanks!

Re: Question About Distributed Mode

Posted by Júlio Pires <ju...@gmail.com>.
Some additional information (now running in debug mode - Hama and
Zookeeper).

However, the examples dont work (again).

I put only the most relevant parts (in my opinion).

--------------------------
In the folder tasklogs:
-----------------------------
13/09/05 10:02:12 DEBUG security.Groups:  Creating new Groups object
13/09/05 10:02:12 DEBUG security.Groups: Group mapping
impl=org.apache.hadoop.security.ShellBasedUnixGroupsMapping;
cacheTimeout=300000
...
The line below as veriquei the forum / jira the hadoop, is harmless..
conf.Configuration: java.io.IOException: config()
...
13/09/05 10:02:36 DEBUG security.UserGroupInformation: hadoop login
13/09/05 10:02:36 DEBUG security.UserGroupInformation: hadoop login commit
13/09/05 10:02:38 DEBUG security.UserGroupInformation: using local
user:UnixPrincipal: hduser
13/09/05 10:02:38 DEBUG security.UserGroupInformation: UGI loginUser:hduser
...
-----------------------------
File bspmaster.log
-----------------------------
2013-09-05 10:00:55,284 INFO org.apache.hama.bsp.JobInProgress: Job is
initialized.
2013-09-05 10:00:55,676 DEBUG org.apache.hama.bsp.Counters: Adding
LAUNCHED_TASKS
2013-09-05 10:01:05,945 DEBUG org.apache.hama.bsp.Counters: Adding
SUPERSTEP_SUM
...
2013-09-05 10:02:15,201 DEBUG org.apache.zookeeper.ClientCnxn: Got ping
response for sessionid: 0x140ed8a99a10001 after 120ms
....
2013-09-05 10:02:46,487 DEBUG org.apache.hama.bsp.Counters: Adding
SUPERSTEP_SUM
2013-09-05 10:02:46,531 INFO org.apache.hama.bsp.JobInProgress: Taskid
'attempt_201309050955_0001_000002_0' has failed.
....
2013-09-05 10:02:46,575 DEBUG org.apache.hama.bsp.JobInProgress: Removing
/tmp/hadoop-hduser/bsp/local/bspMaster/job_201309050955_0001.xml and
/tmp/hadoop-hduser/bsp/local/bspMaster/job_201309050955_0001.jar getJobFile
= hdfs://localhost:54310/tmp/hadoop-hduser/bsp/system/submit_elbm0i/job.xml
2013-09-05 10:02:47,175 INFO org.apache.hama.bsp.JobInProgress: Job failed.
2013-09-05 10:02:47,187 DEBUG org.apache.hama.bsp.JobInProgress: Removing
null and null getJobFile =
hdfs://localhost:54310/tmp/hadoop-hduser/bsp/system/submit_elbm0i/job.xml
2013-09-05 10:02:47,811 DEBUG org.apache.hama.bsp.Counters: Adding
SUPERSTEP_SUM
2013-09-05 10:02:48,531 DEBUG org.apache.hama.bsp.Counters: Adding
SUPERSTEP_SUM
2013-09-05 10:03:35,246 DEBUG org.apache.zookeeper.ClientCnxn: Got ping
response for sessionid: 0x140ed8a99a10001 after 48ms
...
-----------
File zookeeper.out (without errors)
-----------
...2013-09-05 09:56:27,760 [myid:] - INFO  [SyncThread:0:ZooKeeperServer@595]
- Established session 0x140ed8a99a10000 with negotiated timeout 240000 for
client /127.0.0.1:45872
2013-09-05 09:58:36,924 [myid:] - INFO  [NIOServerCxn.Factory:
0.0.0.0/0.0.0.0:21810:NIOServerCnxnFactory@197] - Accepted socket
connection from /127.0.0.1:45876
2013-09-05 09:58:37,120 [myid:] - INFO  [NIOServerCxn.Factory:
0.0.0.0/0.0.0.0:21810:ZooKeeperServer@839] - Client attempting to establish
new session at /127.0.0.1:45876
2013-09-05 09:58:37,272 [myid:] - INFO  [SyncThread:0:ZooKeeperServer@595]
- Established session 0x140ed8a99a10001 with negotiated timeout 240000 for
client /127.0.0.1:45876
...

Some tip of what might be?



2013/9/5 Júlio Pires <ju...@gmail.com>

> In this scenario was not using an external Zookeeper (Hama starts an
> embedded, correct?).
>
> I redid my environment. However, I have found the problem, but not the
> solution.
>
> On physical machines, the example works correctly (localhost/pseudo
> distributed mode).
> In tests on virtual machines via qemu, also works.
>
> But, when I do tests on virtual machines (via cloud) dont work.
>
> Is there any specific configuration to run on virtual machines in Cloud?
>
>
>
> 2013/9/4 Edward J. Yoon <ed...@apache.org>
>
> Please see the zookeeper logs and check whether zookeeper is running.
>> Also, you need to check the "hama.zookeeper.quorum",
>> "hama.zookeeper.property.clientPort" properties are correct.
>>
>>
>>
>> On Wed, Sep 4, 2013 at 10:17 AM, Júlio Pires <ju...@gmail.com>
>> wrote:
>> > Thanks Edward!
>> >
>> > I'm running with the following scenario:
>> > JDK 7
>> > Hadoop: 1.2.1 (is compatible, correct?)
>> > Hama: 0.6.3
>> >
>> > grooservers e  hama.zookeeper.quorum com o endereço das três (master,
>> > slave_a e slave_b).
>> >
>> > However, when trying to run the PI example, the following errors occur:
>> > ------------------------------------------------
>> > ERRO 1 - Arquivo (hama-bspmaster.log)
>> > -------------------------------------------------
>> > 2013-09-04 02:55:39,369 INFO
>> > org.apache.hama.bsp.sync.ZKSyncBSPMasterClient: Initialized ZK false
>> > 2013-09-04 02:55:39,369 INFO org.apache.hama.bsp.sync.ZKSyncClient:
>> > Initializing ZK Sync Client
>> > 2013-09-04 02:55:39,513 *ERROR
>> > org.apache.hama.bsp.sync.ZKSyncBSPMasterClient:
>> > org.apache.zookeeper.KeeperException$ConnectionLossException:
>> > KeeperErrorCode = ConnectionLoss for /bsp*
>> >         at
>> > org.apache.zookeeper.KeeperException.create(KeeperException.java:99)
>> >         at
>> > org.apache.zookeeper.KeeperException.create(KeeperException.java:51)
>> >         at org.apache.zookeeper.ZooKeeper.exists(ZooKeeper.java:1041)
>> >         at org.apache.zookeeper.ZooKeeper.exists(ZooKeeper.java:1069)
>> >         at
>> >
>> org.apache.hama.bsp.sync.ZKSyncBSPMasterClient.init(ZKSyncBSPMasterClient.java:62)
>> >         at org.apache.hama.bsp.BSPMaster.initZK(BSPMaster.java:509)
>> >         at org.apache.hama.bsp.BSPMaster.startMaster(BSPMaster.java:492)
>> >         at org.apache.hama.bsp.BSPMaster.startMaster(BSPMaster.java:475)
>> >         at org.apache.hama.BSPMasterRunner.run(BSPMasterRunner.java:46)
>> >
>> > -----------------------------------------------
>> > ERRO 2: Arquivo: tasklog/attempt_201309040255_0001_000003_0.log"
>> > -----------------------------------------------
>> >
>> > 13/09/04 02:57:21 INFO sync.ZooKeeperSyncClientImpl: Start connecting to
>> > Zookeeper! At /192.168.122.127:61001
>> > 13/09/04 02:57:22 *ERROR sync.ZooKeeperSyncClientImpl:
>> > org.apache.zookeeper.KeeperException$NodeExistsException:
>> KeeperErrorCode =
>> > NodeExists for /bsp/job_201309040255_0001/peers*
>> > 13/09/04 02:57:22 INFO ipc.Server: Starting SocketReader
>> > 13/09/04 02:57:22 INFO ipc.Server: IPC Server Responder: starting
>> > 13/09/04 02:57:22 INFO ipc.Server: IPC Server listener on 61001:
>> starting
>> >
>> > What can it be?
>> >
>> >
>> >
>> > 2013/9/3 Edward J. Yoon <ed...@apache.org>
>> >
>> >> > Questions:
>> >> > 1) In the files "groomservers" and hama-site.xml (property
>> >> > hama.zookeeper.quorum ), should be included the three machines or
>> just
>> >> > slave_a and slave_b?
>> >>
>> >> There's no great difference either way. If the BSP application
>> >> requires large amount of RAM and hdfs-namenode is also started on
>> >> master server, the latter is best (large cluster case).
>> >>
>> >> > 2) With the running environment, how do I add a new slave machine
>> >> > On-the-fly?
>> >>
>> >> Just start groom server to add new slave machine:
>> >>
>> >> user@new_slave_c $ bin/hama-daemons.sh --config $HAMA_CONF_DIR start
>> groom
>> >>
>> >> Then, it will be recognized automatically.
>> >>
>> >> On Wed, Sep 4, 2013 at 3:58 AM, Júlio Pires <ju...@gmail.com>
>> >> wrote:
>> >> > Hello,
>> >> >
>> >> > I have three machines, named as follows: master, slave_a, slave_b
>> >> >
>> >> > Questions:
>> >> > 1) In the files "groomservers" and hama-site.xml (property
>> >> > hama.zookeeper.quorum ), should be included the three machines or
>> just
>> >> > slave_a and slave_b?
>> >> >
>> >> > 2) With the running environment, how do I add a new slave machine
>> >> > On-the-fly?
>> >> >
>> >> > Thanks!
>> >>
>> >>
>> >>
>> >> --
>> >> Best Regards, Edward J. Yoon
>> >> @eddieyoon
>> >>
>>
>>
>>
>> --
>> Best Regards, Edward J. Yoon
>> @eddieyoon
>>
>
>

Re: Question About Distributed Mode

Posted by Júlio Pires <ju...@gmail.com>.
In this scenario was not using an external Zookeeper (Hama starts an
embedded, correct?).

I redid my environment. However, I have found the problem, but not the
solution.

On physical machines, the example works correctly (localhost/pseudo
distributed mode).
In tests on virtual machines via qemu, also works.

But, when I do tests on virtual machines (via cloud) dont work.

Is there any specific configuration to run on virtual machines in Cloud?



2013/9/4 Edward J. Yoon <ed...@apache.org>

> Please see the zookeeper logs and check whether zookeeper is running.
> Also, you need to check the "hama.zookeeper.quorum",
> "hama.zookeeper.property.clientPort" properties are correct.
>
>
>
> On Wed, Sep 4, 2013 at 10:17 AM, Júlio Pires <ju...@gmail.com>
> wrote:
> > Thanks Edward!
> >
> > I'm running with the following scenario:
> > JDK 7
> > Hadoop: 1.2.1 (is compatible, correct?)
> > Hama: 0.6.3
> >
> > grooservers e  hama.zookeeper.quorum com o endereço das três (master,
> > slave_a e slave_b).
> >
> > However, when trying to run the PI example, the following errors occur:
> > ------------------------------------------------
> > ERRO 1 - Arquivo (hama-bspmaster.log)
> > -------------------------------------------------
> > 2013-09-04 02:55:39,369 INFO
> > org.apache.hama.bsp.sync.ZKSyncBSPMasterClient: Initialized ZK false
> > 2013-09-04 02:55:39,369 INFO org.apache.hama.bsp.sync.ZKSyncClient:
> > Initializing ZK Sync Client
> > 2013-09-04 02:55:39,513 *ERROR
> > org.apache.hama.bsp.sync.ZKSyncBSPMasterClient:
> > org.apache.zookeeper.KeeperException$ConnectionLossException:
> > KeeperErrorCode = ConnectionLoss for /bsp*
> >         at
> > org.apache.zookeeper.KeeperException.create(KeeperException.java:99)
> >         at
> > org.apache.zookeeper.KeeperException.create(KeeperException.java:51)
> >         at org.apache.zookeeper.ZooKeeper.exists(ZooKeeper.java:1041)
> >         at org.apache.zookeeper.ZooKeeper.exists(ZooKeeper.java:1069)
> >         at
> >
> org.apache.hama.bsp.sync.ZKSyncBSPMasterClient.init(ZKSyncBSPMasterClient.java:62)
> >         at org.apache.hama.bsp.BSPMaster.initZK(BSPMaster.java:509)
> >         at org.apache.hama.bsp.BSPMaster.startMaster(BSPMaster.java:492)
> >         at org.apache.hama.bsp.BSPMaster.startMaster(BSPMaster.java:475)
> >         at org.apache.hama.BSPMasterRunner.run(BSPMasterRunner.java:46)
> >
> > -----------------------------------------------
> > ERRO 2: Arquivo: tasklog/attempt_201309040255_0001_000003_0.log"
> > -----------------------------------------------
> >
> > 13/09/04 02:57:21 INFO sync.ZooKeeperSyncClientImpl: Start connecting to
> > Zookeeper! At /192.168.122.127:61001
> > 13/09/04 02:57:22 *ERROR sync.ZooKeeperSyncClientImpl:
> > org.apache.zookeeper.KeeperException$NodeExistsException:
> KeeperErrorCode =
> > NodeExists for /bsp/job_201309040255_0001/peers*
> > 13/09/04 02:57:22 INFO ipc.Server: Starting SocketReader
> > 13/09/04 02:57:22 INFO ipc.Server: IPC Server Responder: starting
> > 13/09/04 02:57:22 INFO ipc.Server: IPC Server listener on 61001: starting
> >
> > What can it be?
> >
> >
> >
> > 2013/9/3 Edward J. Yoon <ed...@apache.org>
> >
> >> > Questions:
> >> > 1) In the files "groomservers" and hama-site.xml (property
> >> > hama.zookeeper.quorum ), should be included the three machines or just
> >> > slave_a and slave_b?
> >>
> >> There's no great difference either way. If the BSP application
> >> requires large amount of RAM and hdfs-namenode is also started on
> >> master server, the latter is best (large cluster case).
> >>
> >> > 2) With the running environment, how do I add a new slave machine
> >> > On-the-fly?
> >>
> >> Just start groom server to add new slave machine:
> >>
> >> user@new_slave_c $ bin/hama-daemons.sh --config $HAMA_CONF_DIR start
> groom
> >>
> >> Then, it will be recognized automatically.
> >>
> >> On Wed, Sep 4, 2013 at 3:58 AM, Júlio Pires <ju...@gmail.com>
> >> wrote:
> >> > Hello,
> >> >
> >> > I have three machines, named as follows: master, slave_a, slave_b
> >> >
> >> > Questions:
> >> > 1) In the files "groomservers" and hama-site.xml (property
> >> > hama.zookeeper.quorum ), should be included the three machines or just
> >> > slave_a and slave_b?
> >> >
> >> > 2) With the running environment, how do I add a new slave machine
> >> > On-the-fly?
> >> >
> >> > Thanks!
> >>
> >>
> >>
> >> --
> >> Best Regards, Edward J. Yoon
> >> @eddieyoon
> >>
>
>
>
> --
> Best Regards, Edward J. Yoon
> @eddieyoon
>

Re: Question About Distributed Mode

Posted by "Edward J. Yoon" <ed...@apache.org>.
Please see the zookeeper logs and check whether zookeeper is running.
Also, you need to check the "hama.zookeeper.quorum",
"hama.zookeeper.property.clientPort" properties are correct.



On Wed, Sep 4, 2013 at 10:17 AM, Júlio Pires <ju...@gmail.com> wrote:
> Thanks Edward!
>
> I'm running with the following scenario:
> JDK 7
> Hadoop: 1.2.1 (is compatible, correct?)
> Hama: 0.6.3
>
> grooservers e  hama.zookeeper.quorum com o endereço das três (master,
> slave_a e slave_b).
>
> However, when trying to run the PI example, the following errors occur:
> ------------------------------------------------
> ERRO 1 - Arquivo (hama-bspmaster.log)
> -------------------------------------------------
> 2013-09-04 02:55:39,369 INFO
> org.apache.hama.bsp.sync.ZKSyncBSPMasterClient: Initialized ZK false
> 2013-09-04 02:55:39,369 INFO org.apache.hama.bsp.sync.ZKSyncClient:
> Initializing ZK Sync Client
> 2013-09-04 02:55:39,513 *ERROR
> org.apache.hama.bsp.sync.ZKSyncBSPMasterClient:
> org.apache.zookeeper.KeeperException$ConnectionLossException:
> KeeperErrorCode = ConnectionLoss for /bsp*
>         at
> org.apache.zookeeper.KeeperException.create(KeeperException.java:99)
>         at
> org.apache.zookeeper.KeeperException.create(KeeperException.java:51)
>         at org.apache.zookeeper.ZooKeeper.exists(ZooKeeper.java:1041)
>         at org.apache.zookeeper.ZooKeeper.exists(ZooKeeper.java:1069)
>         at
> org.apache.hama.bsp.sync.ZKSyncBSPMasterClient.init(ZKSyncBSPMasterClient.java:62)
>         at org.apache.hama.bsp.BSPMaster.initZK(BSPMaster.java:509)
>         at org.apache.hama.bsp.BSPMaster.startMaster(BSPMaster.java:492)
>         at org.apache.hama.bsp.BSPMaster.startMaster(BSPMaster.java:475)
>         at org.apache.hama.BSPMasterRunner.run(BSPMasterRunner.java:46)
>
> -----------------------------------------------
> ERRO 2: Arquivo: tasklog/attempt_201309040255_0001_000003_0.log"
> -----------------------------------------------
>
> 13/09/04 02:57:21 INFO sync.ZooKeeperSyncClientImpl: Start connecting to
> Zookeeper! At /192.168.122.127:61001
> 13/09/04 02:57:22 *ERROR sync.ZooKeeperSyncClientImpl:
> org.apache.zookeeper.KeeperException$NodeExistsException: KeeperErrorCode =
> NodeExists for /bsp/job_201309040255_0001/peers*
> 13/09/04 02:57:22 INFO ipc.Server: Starting SocketReader
> 13/09/04 02:57:22 INFO ipc.Server: IPC Server Responder: starting
> 13/09/04 02:57:22 INFO ipc.Server: IPC Server listener on 61001: starting
>
> What can it be?
>
>
>
> 2013/9/3 Edward J. Yoon <ed...@apache.org>
>
>> > Questions:
>> > 1) In the files "groomservers" and hama-site.xml (property
>> > hama.zookeeper.quorum ), should be included the three machines or just
>> > slave_a and slave_b?
>>
>> There's no great difference either way. If the BSP application
>> requires large amount of RAM and hdfs-namenode is also started on
>> master server, the latter is best (large cluster case).
>>
>> > 2) With the running environment, how do I add a new slave machine
>> > On-the-fly?
>>
>> Just start groom server to add new slave machine:
>>
>> user@new_slave_c $ bin/hama-daemons.sh --config $HAMA_CONF_DIR start groom
>>
>> Then, it will be recognized automatically.
>>
>> On Wed, Sep 4, 2013 at 3:58 AM, Júlio Pires <ju...@gmail.com>
>> wrote:
>> > Hello,
>> >
>> > I have three machines, named as follows: master, slave_a, slave_b
>> >
>> > Questions:
>> > 1) In the files "groomservers" and hama-site.xml (property
>> > hama.zookeeper.quorum ), should be included the three machines or just
>> > slave_a and slave_b?
>> >
>> > 2) With the running environment, how do I add a new slave machine
>> > On-the-fly?
>> >
>> > Thanks!
>>
>>
>>
>> --
>> Best Regards, Edward J. Yoon
>> @eddieyoon
>>



-- 
Best Regards, Edward J. Yoon
@eddieyoon

Re: Question About Distributed Mode

Posted by Júlio Pires <ju...@gmail.com>.
Thanks Edward!

I'm running with the following scenario:
JDK 7
Hadoop: 1.2.1 (is compatible, correct?)
Hama: 0.6.3

grooservers e  hama.zookeeper.quorum com o endereço das três (master,
slave_a e slave_b).

However, when trying to run the PI example, the following errors occur:
------------------------------------------------
ERRO 1 - Arquivo (hama-bspmaster.log)
-------------------------------------------------
2013-09-04 02:55:39,369 INFO
org.apache.hama.bsp.sync.ZKSyncBSPMasterClient: Initialized ZK false
2013-09-04 02:55:39,369 INFO org.apache.hama.bsp.sync.ZKSyncClient:
Initializing ZK Sync Client
2013-09-04 02:55:39,513 *ERROR
org.apache.hama.bsp.sync.ZKSyncBSPMasterClient:
org.apache.zookeeper.KeeperException$ConnectionLossException:
KeeperErrorCode = ConnectionLoss for /bsp*
        at
org.apache.zookeeper.KeeperException.create(KeeperException.java:99)
        at
org.apache.zookeeper.KeeperException.create(KeeperException.java:51)
        at org.apache.zookeeper.ZooKeeper.exists(ZooKeeper.java:1041)
        at org.apache.zookeeper.ZooKeeper.exists(ZooKeeper.java:1069)
        at
org.apache.hama.bsp.sync.ZKSyncBSPMasterClient.init(ZKSyncBSPMasterClient.java:62)
        at org.apache.hama.bsp.BSPMaster.initZK(BSPMaster.java:509)
        at org.apache.hama.bsp.BSPMaster.startMaster(BSPMaster.java:492)
        at org.apache.hama.bsp.BSPMaster.startMaster(BSPMaster.java:475)
        at org.apache.hama.BSPMasterRunner.run(BSPMasterRunner.java:46)

-----------------------------------------------
ERRO 2: Arquivo: tasklog/attempt_201309040255_0001_000003_0.log"
-----------------------------------------------

13/09/04 02:57:21 INFO sync.ZooKeeperSyncClientImpl: Start connecting to
Zookeeper! At /192.168.122.127:61001
13/09/04 02:57:22 *ERROR sync.ZooKeeperSyncClientImpl:
org.apache.zookeeper.KeeperException$NodeExistsException: KeeperErrorCode =
NodeExists for /bsp/job_201309040255_0001/peers*
13/09/04 02:57:22 INFO ipc.Server: Starting SocketReader
13/09/04 02:57:22 INFO ipc.Server: IPC Server Responder: starting
13/09/04 02:57:22 INFO ipc.Server: IPC Server listener on 61001: starting

What can it be?



2013/9/3 Edward J. Yoon <ed...@apache.org>

> > Questions:
> > 1) In the files "groomservers" and hama-site.xml (property
> > hama.zookeeper.quorum ), should be included the three machines or just
> > slave_a and slave_b?
>
> There's no great difference either way. If the BSP application
> requires large amount of RAM and hdfs-namenode is also started on
> master server, the latter is best (large cluster case).
>
> > 2) With the running environment, how do I add a new slave machine
> > On-the-fly?
>
> Just start groom server to add new slave machine:
>
> user@new_slave_c $ bin/hama-daemons.sh --config $HAMA_CONF_DIR start groom
>
> Then, it will be recognized automatically.
>
> On Wed, Sep 4, 2013 at 3:58 AM, Júlio Pires <ju...@gmail.com>
> wrote:
> > Hello,
> >
> > I have three machines, named as follows: master, slave_a, slave_b
> >
> > Questions:
> > 1) In the files "groomservers" and hama-site.xml (property
> > hama.zookeeper.quorum ), should be included the three machines or just
> > slave_a and slave_b?
> >
> > 2) With the running environment, how do I add a new slave machine
> > On-the-fly?
> >
> > Thanks!
>
>
>
> --
> Best Regards, Edward J. Yoon
> @eddieyoon
>

Re: Question About Distributed Mode

Posted by "Edward J. Yoon" <ed...@apache.org>.
> Questions:
> 1) In the files "groomservers" and hama-site.xml (property
> hama.zookeeper.quorum ), should be included the three machines or just
> slave_a and slave_b?

There's no great difference either way. If the BSP application
requires large amount of RAM and hdfs-namenode is also started on
master server, the latter is best (large cluster case).

> 2) With the running environment, how do I add a new slave machine
> On-the-fly?

Just start groom server to add new slave machine:

user@new_slave_c $ bin/hama-daemons.sh --config $HAMA_CONF_DIR start groom

Then, it will be recognized automatically.

On Wed, Sep 4, 2013 at 3:58 AM, Júlio Pires <ju...@gmail.com> wrote:
> Hello,
>
> I have three machines, named as follows: master, slave_a, slave_b
>
> Questions:
> 1) In the files "groomservers" and hama-site.xml (property
> hama.zookeeper.quorum ), should be included the three machines or just
> slave_a and slave_b?
>
> 2) With the running environment, how do I add a new slave machine
> On-the-fly?
>
> Thanks!



-- 
Best Regards, Edward J. Yoon
@eddieyoon