[TriLUG] System monitoring tool

[Glen Ford]

Ron Joffe wrote:

>I'm looking for some suggestions for an open source tool (or set of tools) 
>which would allow me to monitor a number of customer systems.
>
>At each customer site I have a number of Linux servers. On each server I 
>currently run a number of shell scripts out of cron for the following 
>processes:
>
>1. Check disk space on given local partitions.
>2. Check multiple types of on disk error logs (these are typically os and 
>application logs which I scan for keywords.
>3. Check multiple application status (i.e. is an oracle process currently 
>running)
>4. Check within oracle database for certain errors (using SQL statements)
>5. Validate status of print queues.
>6. Ping other servers in the network
>
>Currently I have these processes running out of cron on a regular basis 
>(timing depends on a number of factors but can be between every minute to 
>every hour).
>
>If a problem occurs, then I have set up a list of email's to which the system 
>mails the errors.
>
>What I am lacking is a process that allows me to use more of a centralized 
>approach, and a more hierarchy as to the email's that the alerts generate.
>
>For example if disk space is filling up, I would like person #1 to get a 
>single email when it reaches a threshold, and when nobody responds within X 
>minutes and correct s the issue, then send email to person #2 etc.
>
>Also I am looking for a central "Dashboard" to give me an overview of system 
>status. However the client systems would have to connect to my "central 
>server" to update. They would have to push info up, rather then my central 
>server querying them. This is due to network / firewall configurations.
>
>I have looked at sourceforge, etc and have found a number of interesting 
>projects (Zabbix, OpenNMS, OSSIM, etc etc). Does anyone have any experience, 
>suggestions as to which product would fit?
>
>I can spend more time and modify my scripts/code to do this, but just 
>wondering what others are using for similar processes?
>
>Thanks
>
>Ron
>
>
>  
>
BigSister or BigBrother use the push model. They do the standard disk 
and process monitoring. With BigSister you can reference perl scripts to 
monitor d/b.  With BigBrother I think you have to use bourne shell scripts.
The clients send updates over tcp port 1984 to one or more management 
servers. The port is configurable.
The alerts can be sent via email.

/geln


-- 
Glen Ford
gford at idiom.com