[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [hobbit] NFS and shared clientlaunch.cfg files

sup pimp ! nice to see you here.
you idea should work good, it would keep all machines from needlessly starting a script that justs exit. Thing is you'll have to create a clientlaunch.cfg file or simlink to the default one for each machine running hobbit.
Another Idea , similar to Charle Jone's idea, is to launch a wrapper script from clientlaunch.cfg. A wrapper called mycheck just looks for a script named mycheck.$BBHOSTNAME, exec's it if it exists or exits with 0.
more than one way to skin a cat , hobbit's like the swiss army of knife of system monitoring :) 


-Dan

Charles Jones wrote:
I encountered this exact problem, with monitoring a few hundred servers that had a common NFS filesystem. For the most part, Hobbit handles this gracefully as it uses unique names for logfiles and such. I ran into problems though when I wanted an extra script to run, but just on certain hosts. I ended up just having the ext script check the hostname and exit gracefully if it was not the host it should run on: 
HOSTNAME=`/bin/hostname`

if ! echo $HOSTNAME | egrep -q 'app-28|web-12|web-13|db-10' ; then
   exit 0
fi
What I would really like to see (and I posted this on the list before), is a server-side configuration of what scripts or even commands to run on the remote hosts. I doubt this will be implemented though, because too many people view it as a security risk. What I proposed was a config file with a format something like: 

<hostname> <script path> <interval>
example:
host1.domain.com /home/hobbit/client/ext/somescript.sh 5m
The above would instruct the hobbit client on host1.domain.com to run the somescript.sh every 5 minutes.
Another way to do it is just use cron and run the script via bbcmd, but I dont like things that run with hobbit to be able to run when Hobbit is not running (if I have the client down for some reason, the cron would still kickoff unless I also disabled it). 

-Charles


Michael Dunne wrote:


Greetings fellow Hobbit-ers,
I was recently tasked with monitoring 100+ solaris systems. Feeling undaunted I reached into my trusty IT toolbox and pulled out one of my favorite tools, Hobbit. What I encountered is as followed. 
The systems utilize NFS mounted directories. I installed the client files into /usr/local/
hobbit/client and voila! they were available to all of my systems. The issue that I encountered was that I wished to run different external scripts per client, but with a "shared" clientlaunch file I could not see how to do this. (This is where I cringe as I suspect that my solution is rather naive) What I would up doing is modifying the runclient.sh file as reflected below:
~/client hobbit$ diff runclient.sh runclient.ren 79c79 
<               $0 --hostname="$MACHINEDOTS" stop
---

             $0 stop

83c83
<       $HOBBITCLIENTHOME/bin/hobbitlaunch --config=$HOBBITCLIENTHOME/etc/
clientlaunch.cfg --log=$HOBBITCLIENTHOME/logs/clientlaunch.log --pidfile=
$HOBBITCLIENTHOME/logs/clientlaunch.$MACHINEDOTS.pid
---

     $HOBBITCLIENTHOME/bin/hobbitlaunch --config=$HOBBITCLIENTHOME/etc/

$MACHINEDOTS.clientlaunch.cfg --log=$HOBBITCLIENTHOME/logs/
$MACHINEDOTS.clientlaunch.log --pidfile=$HOBBITCLIENTHOME/logs/clientlaunch.
$MACHINEDOTS.pid
102c102
<               $0 --hostname="$MACHINEDOTS" stop
---

             $0 stop

107c107
<       $0 --hostname="$MACHINEDOTS" --os="$BBOSTYPE" start
---

     $0 start
My questions are as followed. 
Has anyone encountered this specific issue before, and if so how did you address it?
Does my, admittedly inelegant, solution pose any issues in relation to long term use? I searched the list and was unable to find anything specific to my issue. 

Thanks in advance for reading this missive.


Best regards,

Mike

To unsubscribe from the hobbit list, send an e-mail to
hobbit-unsubscribe (at) hswn.dk <mailto:hobbit-unsubscribe (at) hswn.dk>