accumulo-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Geoffry Roberts <threadedb...@gmail.com>
Subject Re: Multi Node Question
Date Wed, 17 Sep 2014 19:06:19 GMT
I took a look at the accumulo logs on the non-starting servers.  They were
complaining there was  no route to host, and the problem seemed to be with
port 9000.  I took stab at fixing it by opening said port in iptables.
 That worked in so much as when I did a jps, I saw a Main process running,
so that must be the tserver.  However, with the opening of 9000 hadoop
datanodes would no longer start.  When I closed 9000 (restoring to the
first state) hadoop datanodes started as before and the accumulo nodes quit
starting. it seems to be one or the other.

Was iptables the wrong thing to do? Is there another way?

fwiw, This cluster does not do ssh over port 22.

Thanks

On Wed, Sep 17, 2014 at 9:16 AM, Eric Newton <eric.newton@gmail.com> wrote:

> Yes, you should see a tserver running on every node.
>
> Assuming you are starting everything with the accumulo "start-all.sh"
> script, you will have to add your other nodes to the conf/slaves file.
>
> If you have already done that, there may be something wrong on the other
> hosts (missing files, permissions, ssh keys, etc).
>
> Make sure the accumulo directories sync'd out to every node. Check to make
> sure log directories exist and are writable.  If they are there, go to one
> of those nodes and check for tserver*.out and tserver*.err files, which
> contain the stdout and stderr when the server is started.  Any local issues
> will be in those files.
>
> If the tservers manage to start running, and experience an error, you may
> see it on the monitor pages, which saves you the trouble of ssh'ing to the
> machine and looking in local logs.
>
> -Eric
>
> On Wed, Sep 17, 2014 at 8:46 AM, Geoffry Roberts <threadedblue@gmail.com>
> wrote:
>
>> I have been running a single node Accumulo, but I now have multiple nodes
>> and I have observed that Accumulo's processes only list out (jps) on the
>> node from which I started it.  The other nodes show only Hadoop processes,
>> usually the datanode.
>>
>> Is this as it should be? or should I be seeing Accumulo on all nodes?
>> Other than that, Accumulo seems to be working.
>>
>> Thanks
>>
>> --
>> There are ways and there are ways,
>>
>> Geoffry Roberts
>>
>
>


-- 
There are ways and there are ways,

Geoffry Roberts

Mime
View raw message