Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Plain View
HDFS, mail # user - Re: Node manager health checker opts


Copy link to this message
-
Re: Node manager health checker opts
Tucker 2013-03-26, 05:18
Unfortunately that may not be very feasible.  I'll have to see if there's a
way.  Are my assumptions about how the yarn.nodemanager.health-
checker.script.opts option should work correct?  It took quite a bit of
trial and error to sort this out initially (who knew even an unhealthy node
script run had to have a 0 exit status...) and I'm worried I may be
starting off on the wrong foot here.
On Mon, Mar 25, 2013 at 3:27 PM, Arun C Murthy <[EMAIL PROTECTED]> wrote:

> Hmm... an easy way to debug this would be to add a LOG statement in
> NodeHealthScriptRunner.init to printout the args you are getting from the
> config.
>
> Is there a chance you can try this, recompile & re-run?
>
> thanks,
> Arun
>
> On Mar 25, 2013, at 11:14 AM, Tucker wrote:
>
> Does anyone have a working example of a node manager health checker scipt
> using "yarn.nodemanager.health-checker.script.opts"?  I wrote a health
> checker that works fine but one of the items being checked is a little too
> sensitive.  Since I wrote it to be able to load and unload modules by
> passing various flags.  Unfortunately, adding these flags to my config
> doesn't seem to have had any affect and we've had to disable the health
> check entirely.
>
> For reference:
>
> $ health_checker -h
> Usage: health_checker [options]
>         --default-disabled           Default all checks disabled.
>     -e, --enable-checks CHECKS       Command separated list of checks to
> enable.
>     -d, --disable-checks CHECKS      Command separated list of checks to
> disable.
>     -l, --list                       List available checks.
>
> Settings used:
>
> <property>
> <name>yarn.nodemanager.health-checker.script.path</name>
> <value>/usr/bin/health_checker</value>
> </property>
> ...
> <property>
> <name>yarn.nodemanager.health-checker.script.opts</name>
> <value>-d Network</value>
> </property>
>
> If the flag were actually being passed, I would expect the output to be
> return healthy.  This is what I see on a command line:
>
> # health_checker
> ERROR(s): ["Errors found on interface eth2."]
> # health_checker -d Network
> Healthy
> # echo $?
> 0
>
> Unfortunately, even with opts set, I continue to get the interface errors
> warning after cluster start and beyond the run interval.  I assume I'm
> missing something but I can't seem to find any good docs on the matter.
>
> --
>
> --tucker
>
>
> --
> Arun C. Murthy
> Hortonworks Inc.
> http://hortonworks.com/
>
>
>
--

--tucker