You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@accumulo.apache.org by Aaron Cordova <aa...@cordovas.org> on 2011/12/22 19:14:20 UTC
init new instance on EC2
I'm setting up a new instance of Accumulo 1.3.5, using the binary distribution available on the downloads page, on EC2, using Cloudera's distribution of hadoop and zookeeper, and after initialization, upon trying to start the master get this error:
ERROR: Problem getting real goal state: org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode = NoNode for /accumulo/68ce2daf-57ee-401f-91c6-dc0df0fcaac9/masters/goal_state
which repeats endlessly. Is the master not creating the right nodes under zookeeper? This happens when using a cluster of three zookeeper servers as well as when using a single stand-alone, local zookeeper server. Also, I'm starting Accumulo thus:
bin/accumulo logger &
bin/accumulo tserver &
bin/accumulo master &
Could that be the source of the problem? Does one of the startup scripts create the proper master nodes? bin/start-here.sh seems to have no effect - i.e. no services are started.
Aaron
Re: init new instance on EC2
Posted by Aaron Cordova <aa...@cordovas.org>.
Thanks Minh - that set seems to be working for me.
On Dec 22, 2011, at 1:40 PM, Minh Nguyen wrote:
> The developers should confirm this, but these are the ports that I know about:
>
> gc.port.client = 50091
> logger.port.client = 11224
> master.port.client = 9999
> monitor.port.client = 50095
> monitor.port.log4j = 4560
> trace.port.client = 12234
> tserver.port.client = 9997
>
> ~ Minh
>
> On Thu, Dec 22, 2011 at 1:36 PM, Aaron Cordova <aa...@cordovas.org> wrote:
>
Re: init new instance on EC2
Posted by Minh Nguyen <mn...@bericotechnologies.com>.
*The developers should confirm this, but these are the ports that I know
about:*
*
*
*gc.port.client = 50091
logger.port.client = 11224
master.port.client = 9999
monitor.port.client = 50095
monitor.port.log4j = 4560
trace.port.client = 12234
tserver.port.client = 9997*
*
*
* ~ Minh*
On Thu, Dec 22, 2011 at 1:36 PM, Aaron Cordova <aa...@cordovas.org> wrote:
> Also - is there a list of port numbers that need to be open for
> interprocess comms?
>
> On Dec 22, 2011, at 1:28 PM, John W Vines wrote:
>
> > The bin/start-all.sh script seeds the initial goal state for the master
> as of 1.4. Goal state is what drives the actions of the master.
> >
> > John
> >
> > ----- Original Message -----
> > | From: "Aaron Cordova" <aa...@cordovas.org>
> > | To: accumulo-user@incubator.apache.org
> > | Sent: Thursday, December 22, 2011 1:14:20 PM
> > | Subject: init new instance on EC2
> > | I'm setting up a new instance of Accumulo 1.3.5, using the binary
> > | distribution available on the downloads page, on EC2, using Cloudera's
> > | distribution of hadoop and zookeeper, and after initialization, upon
> > | trying to start the master get this error:
> > |
> > | ERROR: Problem getting real goal state:
> > | org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode
> > | = NoNode for
> > | /accumulo/68ce2daf-57ee-401f-91c6-dc0df0fcaac9/masters/goal_state
> > |
> > | which repeats endlessly. Is the master not creating the right nodes
> > | under zookeeper? This happens when using a cluster of three zookeeper
> > | servers as well as when using a single stand-alone, local zookeeper
> > | server. Also, I'm starting Accumulo thus:
> > |
> > | bin/accumulo logger &
> > | bin/accumulo tserver &
> > | bin/accumulo master &
> > |
> > | Could that be the source of the problem? Does one of the startup
> > | scripts create the proper master nodes? bin/start-here.sh seems to
> > | have no effect - i.e. no services are started.
> > |
> > | Aaron
>
>
Re: init new instance on EC2
Posted by Aaron Cordova <aa...@cordovas.org>.
Also - is there a list of port numbers that need to be open for interprocess comms?
On Dec 22, 2011, at 1:28 PM, John W Vines wrote:
> The bin/start-all.sh script seeds the initial goal state for the master as of 1.4. Goal state is what drives the actions of the master.
>
> John
>
> ----- Original Message -----
> | From: "Aaron Cordova" <aa...@cordovas.org>
> | To: accumulo-user@incubator.apache.org
> | Sent: Thursday, December 22, 2011 1:14:20 PM
> | Subject: init new instance on EC2
> | I'm setting up a new instance of Accumulo 1.3.5, using the binary
> | distribution available on the downloads page, on EC2, using Cloudera's
> | distribution of hadoop and zookeeper, and after initialization, upon
> | trying to start the master get this error:
> |
> | ERROR: Problem getting real goal state:
> | org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode
> | = NoNode for
> | /accumulo/68ce2daf-57ee-401f-91c6-dc0df0fcaac9/masters/goal_state
> |
> | which repeats endlessly. Is the master not creating the right nodes
> | under zookeeper? This happens when using a cluster of three zookeeper
> | servers as well as when using a single stand-alone, local zookeeper
> | server. Also, I'm starting Accumulo thus:
> |
> | bin/accumulo logger &
> | bin/accumulo tserver &
> | bin/accumulo master &
> |
> | Could that be the source of the problem? Does one of the startup
> | scripts create the proper master nodes? bin/start-here.sh seems to
> | have no effect - i.e. no services are started.
> |
> | Aaron
Re: init new instance on EC2
Posted by Adam Fuchs <ad...@ugov.gov>.
Aaron,
Somehow that SetGoalState line didn't make it into 1.3.5 (probably lost
sometime in the export), but the fix is in for 1.4 and later. There's only
the one line different in bin/start-here.sh in the two branches, so you can
copy the 1.4 start-here.sh over to your instance and it should work for you.
Adam
On Thu, Dec 22, 2011 at 1:56 PM, Aaron Cordova <aa...@cordovas.org> wrote:
> Correction to my earlier email - it appears running start-all.sh, not
> start-here.sh necessary to set the goal state for 1.3.5.
>
> On Dec 22, 2011, at 1:28 PM, John W Vines wrote:
>
> > The bin/start-all.sh script seeds the initial goal state for the master
> as of 1.4. Goal state is what drives the actions of the master.
> >
> > John
>
>
Re: init new instance on EC2
Posted by Aaron Cordova <aa...@cordovas.org>.
Correction to my earlier email - it appears running start-all.sh, not start-here.sh necessary to set the goal state for 1.3.5.
On Dec 22, 2011, at 1:28 PM, John W Vines wrote:
> The bin/start-all.sh script seeds the initial goal state for the master as of 1.4. Goal state is what drives the actions of the master.
>
> John
Re: init new instance on EC2
Posted by Aaron Cordova <aa...@cordovas.org>.
It appears that one must use start-here.sh and ssh, even into localhost, in order for the goal state to be set. So bin/accumulo master won't work on it's own.
On Dec 22, 2011, at 1:28 PM, John W Vines wrote:
> The bin/start-all.sh script seeds the initial goal state for the master as of 1.4. Goal state is what drives the actions of the master.
>
> John
>
> ----- Original Message -----
> | From: "Aaron Cordova" <aa...@cordovas.org>
> | To: accumulo-user@incubator.apache.org
> | Sent: Thursday, December 22, 2011 1:14:20 PM
> | Subject: init new instance on EC2
> | I'm setting up a new instance of Accumulo 1.3.5, using the binary
> | distribution available on the downloads page, on EC2, using Cloudera's
> | distribution of hadoop and zookeeper, and after initialization, upon
> | trying to start the master get this error:
> |
> | ERROR: Problem getting real goal state:
> | org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode
> | = NoNode for
> | /accumulo/68ce2daf-57ee-401f-91c6-dc0df0fcaac9/masters/goal_state
> |
> | which repeats endlessly. Is the master not creating the right nodes
> | under zookeeper? This happens when using a cluster of three zookeeper
> | servers as well as when using a single stand-alone, local zookeeper
> | server. Also, I'm starting Accumulo thus:
> |
> | bin/accumulo logger &
> | bin/accumulo tserver &
> | bin/accumulo master &
> |
> | Could that be the source of the problem? Does one of the startup
> | scripts create the proper master nodes? bin/start-here.sh seems to
> | have no effect - i.e. no services are started.
> |
> | Aaron
Re: init new instance on EC2
Posted by John W Vines <jo...@ugov.gov>.
The bin/start-all.sh script seeds the initial goal state for the master as of 1.4. Goal state is what drives the actions of the master.
John
----- Original Message -----
| From: "Aaron Cordova" <aa...@cordovas.org>
| To: accumulo-user@incubator.apache.org
| Sent: Thursday, December 22, 2011 1:14:20 PM
| Subject: init new instance on EC2
| I'm setting up a new instance of Accumulo 1.3.5, using the binary
| distribution available on the downloads page, on EC2, using Cloudera's
| distribution of hadoop and zookeeper, and after initialization, upon
| trying to start the master get this error:
|
| ERROR: Problem getting real goal state:
| org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode
| = NoNode for
| /accumulo/68ce2daf-57ee-401f-91c6-dc0df0fcaac9/masters/goal_state
|
| which repeats endlessly. Is the master not creating the right nodes
| under zookeeper? This happens when using a cluster of three zookeeper
| servers as well as when using a single stand-alone, local zookeeper
| server. Also, I'm starting Accumulo thus:
|
| bin/accumulo logger &
| bin/accumulo tserver &
| bin/accumulo master &
|
| Could that be the source of the problem? Does one of the startup
| scripts create the proper master nodes? bin/start-here.sh seems to
| have no effect - i.e. no services are started.
|
| Aaron