Problem deploying bundle data-analytics-with-sql-like

[Ken Williams]

Thanks lots, Sam!!

This is all very, very helpful advice.

Best Regards,

Ken


On 5 February 2015 at 14:59, Samuel Cozannet <samuel.cozannet at canonical.com>
wrote:

> Hey Ken,
>
> My advice if you fall into an error state with a bundle like this is to
> deploy it charm by charm (or use the deployment script provided), as those
> may manage race conditions between the charms.
>
> Let me explain this a little more what we call a race condition.
> When deploying complex solutions, it happens that several charms try to
> "change" a relation at the same time. If setting the relation takes too
> long, it sometimes fails the next relation changes.
> In short, there is a race between 2 services trying to access the same
> resource at the same time and that specific resource can only be consumed
> once at a time.
> This is what is happening to you.
>
>
> How do we fix that?
> * First of all, we have a juju-deployer for bundles that is a lot more
> intelligent as the default deployment for bundles (
> http://pythonhosted.org/juju-deployer/config.html). I encourage you to
> use this for bundles
> * You can also retry the hook by typing "juju resolved -r yarn-hdfs-master/0"
> which will tell Juju you fixed the issue and it should retry the previously
> failed hook.
> * If you're not familiar with the workload and a service broke, maybe the
> old "reboot" trick will fix your issue... "juju run --unit yarn-hdfs-master/0
> sudo reboot" will do... Not guaranteed to work but...
> * In the end, this is all because of the charm not managing the status of
> his fellow charm in a relation. You can fill a bug vs. that charm and
> explain under which conditions you saw the issues, and/or you can also hack
> into the code yourself and try to fix it. In your case here, the problem is
> probably that the service is still down while trying to join. You may check
> the service port is up and running, and wait until it is before actually
> firing the hook. (This is a basic but common example).
>
> Hope it helps!
> Best,
> Samuel
>
>
>
>
>
> Best,
> Samuel
>
> --
> Samuel Cozannet
> Cloud, Big Data and IoT Strategy Team
> Business Development - Cloud and ISV Ecosystem
> Changing the Future of Cloud
> Ubuntu <http://ubuntu.com>  / Canonical UK LTD <http://canonical.com> /
> Juju <https://jujucharms.com>
> samuel.cozannet at canonical.com
> mob: +33 616 702 389
> skype: samnco
> Twitter: @SaMnCo_23
>
> On Thu, Feb 5, 2015 at 1:55 PM, Ken Williams <ken.w at theasi.co> wrote:
>
>>
>> Hi all,
>>
>>    I'm currently using the 'data-analytics-with-sql-like' bundle.
>>    Sometimes it works fine but sometimes it deploys with an error
>>    and I don't know how to fix it without doing 'destroy-environment' and
>>    deploy again (which takes time).
>>
>>    The error (in 'juju stat') is,
>>
>>       yarn-hdfs-master/0:
>>         agent-state: error
>>         agent-state-info: 'hook failed: "namenode-relation-joined" for
>> compute-node:datanode'
>>         agent-version: 1.21.1
>>         machine: "4"
>>
>>
>>     When this error occurs, I can ssh onto yarn-hdfs-master but
>>     I cannot 'hdfs dfs -put' any data onto hdfs.
>>
>>     Is there any way I can fix this without destroying the environment
>>     and deploying again?
>>
>>     Thankyou for any help, please,
>>
>>
>> Ken
>>
>>
>>
>> --
>> Juju mailing list
>> Juju at lists.ubuntu.com
>> Modify settings or unsubscribe at:
>> https://lists.ubuntu.com/mailman/listinfo/juju
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/juju/attachments/20150205/5ffced7f/attachment.html>