deployer hangs when instances are 'locked in pending'

[Charles Butler]

⟫ juju deployer -e hp-test -c bundles.yaml
2014-11-03 16:02:53 Using deployment hdp-hadoop-hive-mysql
2014-11-03 16:02:53 Starting deployment of hdp-hadoop-hive-mysql
2014-11-03 16:02:54 Deploying services...
2014-11-03 16:02:55 Deploying service compute-node using cs:trusty/hdp-hadoop-3
2014-11-03 16:03:06 Deploying service hdphive using cs:trusty/hdp-hive-1
2014-11-03 16:03:17 Deploying service mysql using cs:trusty/mysql-4
2014-11-03 16:03:29 Deploying service yarn-hdfs-master using cs:trusty/hdp-hadoop-3

the response about 20 minutes in:

  "1":
    agent-state-info: no instance types in region-a.geo-1 matching constraints "arch=amd64
      mem=12288M root-disk=40960M"
    instance-id: pending
    series: trusty
  "2":
    agent-state-info: no instance types in region-a.geo-1 matching constraints "arch=amd64
      mem=12288M root-disk=40960M"
    instance-id: pending
    series: trusty

It seems that it would be intuitive for juju deployer to either grep the status output occasionally to see if the deployment has failed for one or more nodes being stuck in pending, or somehow adjust for this behavior - as juju itself will sit in a pending state with this machine never actually deployed - and deployer is driving on autopilot in no particular direction.

dpkg -l | grep juju-deployer
ii juju-deployer 0.4.1-1

[David Britton (dpb)]

Hi Charles --

Does the exiting timeout option not work for you?

  -t TIMEOUT, --timeout TIMEOUT
                        Timeout (sec) for entire deployment (45min default)

On Mon, Nov 3, 2014 at 2:33 PM, Charles Butler <<email address hidden>
> wrote:

> Public bug reported:
>
> ⟫ juju deployer -e hp-test -c bundles.yaml
> 2014-11-03 16:02:53 Using deployment hdp-hadoop-hive-mysql
> 2014-11-03 16:02:53 Starting deployment of hdp-hadoop-hive-mysql
> 2014-11-03 16:02:54 Deploying services...
> 2014-11-03 16:02:55 Deploying service compute-node using
> cs:trusty/hdp-hadoop-3
> 2014-11-03 16:03:06 Deploying service hdphive using cs:trusty/hdp-hive-1
> 2014-11-03 16:03:17 Deploying service mysql using cs:trusty/mysql-4
> 2014-11-03 16:03:29 Deploying service yarn-hdfs-master using
> cs:trusty/hdp-hadoop-3
>
>
> the response about 20 minutes in:
>
> "1":
> agent-state-info: no instance types in region-a.geo-1 matching
> constraints "arch=amd64
> mem=12288M root-disk=40960M"
> instance-id: pending
> series: trusty
> "2":
> agent-state-info: no instance types in region-a.geo-1 matching
> constraints "arch=amd64
> mem=12288M root-disk=40960M"
> instance-id: pending
> series: trusty
>
>
> It seems that it would be intuitive for juju deployer to either grep the
> status output occasionally to see if the deployment has failed for one or
> more nodes being stuck in pending, or somehow adjust for this behavior - as
> juju itself will sit in a pending state with this machine never actually
> deployed - and deployer is driving on autopilot in no particular direction.
>
> dpkg -l | grep juju-deployer
> ii juju-deployer 0.4.1-1
>
> ** Affects: juju-deployer
> Importance: Undecided
> Status: New
>
> --
> You received this bug notification because you are subscribed to juju-
> deployer.
> https://bugs.launchpad.net/bugs/1388990
>
> Title:
> deployer hangs when instances are 'locked in pending'
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/juju-deployer/+bug/1388990/+subscriptions
>

--
David Britton <email address hidden>