accumulo-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Eric Newton <eric.new...@gmail.com>
Subject Re: start-all.sh should do some basic verification (was Accumulo fails to start)
Date Fri, 16 Mar 2012 16:21:21 GMT
We can check for registrations in zookeeper, but this might take as long as
the zookeeper timeout (30 seconds, by default), especially in a restart
situation.

-Eric

On Fri, Mar 16, 2012 at 12:08 PM, Robert Vesse <rvesse@yarcdata.com> wrote:

>  Hi John
>
>  A more robust way to detect JAVA_HOME would be nice but I don't have any
> great ideas on that front
>
>  What I was thinking was more along the lines of can something in
> start-all.sh ping the servers that should have started as a basic sanity
> check to see that Accumulo actually started up?
>
>  Rob
>
>  On Mar 15, 2012, at 2:31 PM, John Vines wrote:
>
> In the debian scripts, I attempt to find JAVA_HOME by tracing
> /usr/bin/java with readlink, but this doesn't exist on OSX and readlink
> behaves differently on RHEL. So then I changed it to just iterating over a
> list of known java directories, which it checks for until it works. I would
> love to have a more standard solution for that as well, if anyone can think
> of one.
>
> On Thu, Mar 15, 2012 at 6:37 PM, Robert Vesse <rvesse@yarcdata.com> wrote:
>
>>  Hi All
>>
>>  This is an RFE based on my troubles getting started with Accumulo which
>> I feel would benefit other users.
>>
>>  As seen in my original thread it is possible to end up with a
>> configuration where start-all.sh will appear to work fine but Accumulo will
>> actually not have started and no logs will have been generated.  This is
>> very unhelpful to the user since they have no visibility that something has
>> gone wrong
>>
>>  Would it be possible for the start-all script to do some basic
>> verification to check that servers actually started up ok and if not print
>> an error "Accumulo did not start correctly, please check your configuration"
>>
>>  Rob
>>
>>  On Mar 14, 2012, at 5:39 PM, Eric Newton wrote:
>>
>> Well, JAVA_HOME expects a certain file layout... for example, on my mac
>> laptop, /Library/Java/Home works, but on Linux it's completely different.
>>
>>  It is especially sensitive when building the JNI libs for the native
>> map.
>>
>>  Since it is hard to debug different configurations on different
>> platforms, it would be great if you could document what works for you.
>>
>>  -Eric
>>
>> On Wed, Mar 14, 2012 at 4:05 PM, Robert Vesse <rvesse@yarcdata.com>wrote:
>>
>>> I have tried to set it in that script but the scripts don't seem to
>>> accept it - this is further confused by the fact that a couple of days ago
>>> everything worked without any issues and I haven't changed my config at all
>>>
>>>  A which java on my command line will give /usr/bin/java as the path
>>> but setting JAVA_HOME to this causes scripts to fail completely because
>>> they blindly append /bin/java to the path, yet bizarrely setting it to /usr
>>> gives the warning seen in my email
>>>
>>>  I will experiment and see whether this is the cause of the issue but
>>> should there not be a way to actually show a useful error message to the
>>> user in this case?
>>>
>>>  Rob
>>>
>>>  On Mar 14, 2012, at 12:56 PM, Eric Newton wrote:
>>>
>>> Are you setting JAVA_HOME in conf/accumulo-env.sh ?
>>>
>>>  If you are setting configuration in your environment, it may not
>>> survive an ssh to localhost.
>>>
>>>  -Eric
>>>
>>> On Wed, Mar 14, 2012 at 3:51 PM, Robert Vesse <rvesse@yarcdata.com>wrote:
>>>
>>>> I'm having problems with Accumulo not starting correctly - yet from the
>>>> command line it looks to start fine i.e. when I run bin/start-all.sh I see
>>>> the following output:
>>>>
>>>> Starting tablet servers and loggers .... done
>>>> Unable to find a $JAVA_HOME at "/usr", continuing with system-provided
>>>> Java...
>>>> Unable to find a $JAVA_HOME at "/usr", continuing with system-provided
>>>> Java...
>>>> Starting logger on localhost
>>>> Starting tablet server on localhost
>>>> Starting master on localhost
>>>> Starting garbage collector on localhost
>>>> Starting monitor on localhost
>>>> Starting tracer on localhost
>>>>
>>>> Yet if I try to start an accumulo shell with bin/accumulo shell -u root
>>>> I get the following error:
>>>>
>>>> 14 12:41:27,472 [impl.ServerClient] WARN : Failed to find an available
>>>> server in the list of servers: []
>>>> 14 12:41:27,473 [shell.Shell] ERROR:
>>>> org.apache.accumulo.core.client.AccumuloException:
>>>> org.apache.thrift.transport.TTransportException: Failed to connect to a
>>>> server
>>>>
>>>> I looked under the logs/ directory and see that a bunch of logs are
>>>> created but they are all zero bytes in size so no helpful hints there :-(
>>>>
>>>> I have followed the README to configure everything appropriately for a
>>>> single node setup (I have Hadoop and Zookeeper running fine) and I actually
>>>> had Accumulo working correctly a couple of days ago yet now I am unable to
>>>> start things up again.  I have tried reformatting my HDFS (as this is just
>>>> a dev setup) and starting from scratch but I consistently hit this issue
>>>> that Accumulo just does not start properly.
>>>>
>>>> Any help or guidance on this would be much appreciated
>>>>
>>>> Rob
>>>
>>>
>>>
>>>
>>
>>
>
>

Mime
View raw message