Server Monitoring - Munin Vs. MRTG - Any Recommendations?

V

Vibe

Verified User
Joined
Aug 3, 2005
Messages
120
Hi All!

I have recently installed Munin as a monitoring tool for one of our servers (in house testing). From what I understand it produces pretty much the same graphs as MRTG. However, it does so using Perl/Shell scripts.

The one thing that I'm NOT a fan of is that every 5 minutes Perl runs (a la Munin Cron) and utilizes some fairly high CPU resources. The reason being is that I am currently monitoring the following:

Apache accesses
Apache activity
Apache processes
Apache volume
Filesystem usage (in %)
Inode usage (in %)
Exim Mailqueue
Exim mail throughput
ProFTP Server Bytes
ProFTP Server Transfers
MySQL throughput
MySQL Connections
MySQL queries
MySQL slow queries
MySQL threads
bge0 traffic
bge1 traffic
bge0 Errors & Collisions
bge1 Errors & Collisions
Netstat
Dovcecot Logins
Number of Processes
VMstat
Auth Log Parser
CPU usage
Load average
Memory usage
File table usage
Swap in/out
Uptime

Can anyone help with a few questions?

(1) From my understanding MRTG utilizes SNMP - Does SNMP consume less memory or processor usage when running? (e.g. unlike the spikes I see with Munin every 5 minutes)

(2) I have searched to see whether MRTG can capture the same information (listed above) - offhand, can MRTG capture information for EXIM, Dovecot and ProFTP?

Thanks for your assistance!
 
So...I take it nobody uses either Munin or MRTG to monitor their servers?

:rolleyes:
 
I currently use Munin and used MRTG a few years ago. Munin seems less mature and more heavy, but I love how it works.
I don't know anything about how is MRTG today, but I know you can do more or less anything (not only by SNMP), it's just more difficult to setup. I remember that sometimes it was easier to use rrdtools directly with custom scripts.
 
IMHO Cacti is great if you want to personalize your graphs completely, while Nagios has a few interesting plugins if you want centralized complete statistics on a large network, but both of them are too big (and eat too many resources) for simple statistics on a single server.
 
A little off topic but related... I only have a couple boxes online ATM (small company) so I use ubersmith to do snmp polling on my boxes. If your datacenter has it, it's a great option.

I manage a large government network that spans several physical locations and we use OpManager to monitor servers and devices at all of the sites. Depending on the size of your orgainization you'll need a VPS or dedicated box to run it, but it's one of the best products I've seen. They have a free edition that monitors up to 10 devices.
http://www.manageengine.com/products/opmanager/index.html
 
I use munin and collect pretty much the same set of info. If you setup the graphs to be generated on demand (via cgi), that reduces the processor load significantly. On the other hand with on-demand generation, you may need to wait as much as a minute to generate all the graphs on the node page. I find that acceptable for me, you may feel differently.
 
You must log in or register to reply here.
Back
Top