Service status shows as "Unknown Health" for custom parcel for Apache Drill

[#1]

I have created a parcel and custom service descriptor (CSD) for Apache Drill but the service status will only report “Unknown Health” (question mark icon) when the Drill service is run from Cloudera Manager v5.4.3. No graphs will display on the Charts portion of the UI. Both “Informational Events” or “Important Events and Alerts” say “NO DATA”.

What I’ve done:

I modified the Apache Drill start script to make it compatible with supervisord as indicated by https://github.com/cloudera/cm_ext/wiki/Control-Scripts.  The service is now able to be started and stopped in a compatible way and the stdout.log and stderr.log in /var/run/cloudera-scm-agent/process/<NNNN>-drill-DRILLBIT_SERVER/logs/ support this.

Basic sanity testing has confirmed that Drill is working.

When the service is run, the status of the service will only show as "Unknown Health" in the Cloudera Manager Web UI.  Other roles (HDFS Datanode) running on the same cluster node are correctly reporting their status. Since supervisord is used to run, and I’m assuming monitor, the services it starts I used /usr/lib64/cmf/agent/build/env/bin/supervisorctl to see what it was reporting. According to it, drill is happily running which it is.

e.g.

$ cd /var/run/cloudera-scm-agent/supervisor

$ /usr/lib64/cmf/agent/build/env/bin/supervisorctl

5380-hdfs-DATANODE        RUNNING     pid 23842, updtime 6 days, 2:59:06

5393-drill-DRILLBIT_SERVER       RUNNING pid 28750, uptime 1 day, 23:38:34

cmflistener      RUNNING     pid 26151, uptime 9 days, 3:01:01

$ ps –ef | grep drill

root      19265  16618  0          19:29   pts/1    00:00:00          grep drill

drill      28750  26150  0          Jan25   ?          00:03:31          /usr/java/default/jre/bin/java –Dlog.path=/opt/cloudera/parcels/APACHE_DRILL-1.4.0/log/drillbit.log …

So, is there something else which is reporting the status of the service to the Cloudera Manager Web UI? Any suggestions on what I should be looking at to fix this problem? 

I appreciate any advice you can offer.

Thanks.

[#2 · Accepted solution]

I didn't receive any feedback, but I thought I'd give an update...

I ended up deploying the custom parcel & csd onto a hardware based cluster, as opposed to the vm based cluster I was originally using, and the service status was correctly being reported as "Good Health".  I'm going to assume it was something related to running in the vm environment that caused the issue as a result.

[#3]

Hey Liam, any interest in sharing your Drill CSD with the world? We'd love to use it and contribute back!