ROOTVG

Hi

Currently we are using Ganglia (http://ganglia.sourceforge.net/) with M. Perzls Power 5 extensions (http://perzl.org/) to monitor our systems. We would like to give CACTI (http://cacti.net/) a try and now we need a advice how to get power 5 / power 6 extensions into CACTI. CACTI is based on SNMP and may be more comfortable to set up than installing gmonds and gmetads on the systems.
I'm pretty new to the SNMP thing but i know that there are mibs and configs to get the power 5 feature - values into SNMP but I don't know the limitations neiter do I know how to set this up.
Our environment is a mixed bunch (about 150 lpars) of power 4, power 5 and power 6  Hardware with AIX 5.2/5.3/6.1 installed. How about integrating the HMC into CACTI, we would be very happy when we could see an "over-all" CPU performance of the whole Regatta/Squadron/Eclipse Frame to be able to do some "predictions" about HW planing.
Is there somebody who can give me some advices? Anyone else tried this before?

Thanks in advance!
Sebastian

I think there was some discussion earlier in the forums - but did not come to a complete resolution. When I get finished with spamassassin - I'll look into it.

Please keep us posted on your learning experiences!

Thanks, that would be nice. I only found a somehow related one with HACMP monitoring over CACTI.
Cheers Sebastian 

Well, run into a snag with spamassassin. I have downloaded the source files, etc. and now I need to do some reading. And look more closely at yor initial question :P

I run Cacti and monitor 2 p650s with 4 LPARS each and 1 p550 with 4 LPARS and several F50's all with Cacti.  AIX 5.2.x.x has weird SNMP issues with Cacti, like CPU's always reporting 100% utilization.  On AIX 5.3.7.0 however I have had no problems at all getting stats on the following.

Multiple Nics at 10\100\1000 no 10Gig yet
CPU's
Disk Space by Filesystem - Total available and ammount used.
Swap
Real Memory
Paging

I pull our Fiber Channel traffic from the switch, but I was able to get those stats from my AIX boxes but not as accurately as from the switch.

I will try to answer questions if I can. 

Thanks kondoor - I was getting nowhere.

Kondoor,
  What did you have to change to get SNMP to report the CPU and interface information? I have been banging my head against the wall for almost 3 weeks now trying to get these stats.

My setups are on p5 550s running AIX 05.03.0000.0054 and using both the stock SNMP daemon on one server and Net-SNMP on the other.

Any tips would be appreciated as I cannot pull these values from either.

Thanks

I am using snmp deamon that comes with AIX on 5.3.x, I had to use Net-SNMP on AIX 5.2.x in order to get anything to report.

I am using Cacti 0.8.7a

I am using templates I found on the Cacti forums http://forums.cacti.net/about6072.html, but I can also use the generic snmp-enabled host templates in cacti to monitor everything.

Make sure you have /etc/snmpd.conf or /etc/snmpdv3.conf configured I don't use V3 so in my configs i use snmpd.conf

Make sure all of your snmpd services are started
snmpd, snmpmibd, hostmibd, and aixmibd

I am reading that you can pull stats on some things just not cpu and net interface is that correct?

Can you walk the tree in command line and get the interface stats?

On the server that only has the builtin snmpd installed I can walk 1.3.6.1.2.1.4.20.1.1 and pull all of the interface stats. This one is working corectly. However on the other server with the latest version of net-snmp I cannot pull anything for that OID. Both of the servers use the same configuration file (Copyed off of the one that works)

I have tried setting up proxying from net-snmp to the AIX snmpd server (Running on port 1610). Even with stopping the net-snmp daemon I cannot pull the interface stats from this host.

Locally I can pull some information using "snmpinfo -m dump" but I cannot get it to be walkable from another host at that port. Is there some permissions that need to be set to allow me to walk this host remotely?

Thanks

I had some problems when i upgraded a server from aix 5.2 to 5.3 and tried to stop using net-snmpd and use the built in snmpd.  Not sure if this will help but worth a try.  stop net-snmp and all the associated services, snmpmibd, hostmibd, snmpd, and aixmibd.  Then restart the services and not net-snmp.  If its an option I might even take net-snmpd out of inittab or rc.d or wherever you are starting it from and restart the server. Other things I have found as gotchas, snmp v1 works great, snmp v2 has issues, snmp v3 I have never gotten to work, so that said I always just use snmp v1.  I don't use passwords and I keep everything set to the default public group for read access.  Below is what the snmpd.conf looks like on all 16 lpars i monitor.  Assuming you are on 5.3 or higher you should not need to use net-snmpd

logging         file=/usr/tmp/snmpd.log         enabled
logging         size=100000                     level=0

community       public
#community       private 127.0.0.1 255.255.255.255 readWrite
#community       system  127.0.0.1 255.255.255.255 readWrite 1.17.2

view            1.17.2          system enterprises view

trap            public          127.0.0.1       1.2.3   fe      #loopback

#snmpd          maxpacket=1024 querytimeout=120 smuxtimeout=60

smux            1.3.6.1.4.1.2.3.1.2.1.2         gated_password  #gated
smux            1.3.6.1.4.1.2.3.1.2.2.1.1.2     dpid_password   #dpid


snmpd smuxtimeout=200 #muxatmd
smux 1.3.6.1.4.1.2.3.1.2.3.1.1 muxatmd_password #muxatmd

I tried dropping net-snmp from the host in question but in doing so it took that host out of monitoring. The monitoring tool that we use is Zenoss and we abuse its trending information. If I switch back to the AIX snmpd I can get all the interface stats that I want into Zenoss, But I loose the CPU and ram monitoring with it.

I am thinking that I need to proxy some snmp queries to the AIX snmpd but have been unable to do so. Unfortunently, Zenoss is way to intergrated with the rest of the network to start replacing it now, and the boss wants one tool to use to get everything monitored.

Do you know how I would go about getting net-snmp to proxy to the builting snmpd correctly? The only OIDs that I need to monitor are
1.3.6.1.2.1.4.20.1.1
1.3.6.1.2.1.4.21.1.1
if that helps any.

Otherwise, any good sugestions for a multi-os monitoring tool that is as non-invasive as snmp?


Thanks

I dont have much net-snmp proxy setup experience.  Just reading thru some docs and web posts this would be kind of what you want to do.  it gets longer when you start using V2 or V3 as you need to include your auth in the proxy line.

proxy -v 1 -c remote_host .1.3.6.1.2.1.4.

com2sec readonly default public

http://net-snmp.sourceforge.net/wiki/index.php/Snmpd_proxy

Wish I could be of more help

Thanks for the help. I am starting to suspect that our vendor build net-snmp with some odd options (unfortunently they come preconfigured to "help" us out) and I will try to rebuild it correctly with clean sources. Thanks for the help and I will update here if I get it working.

I solved the issue that we were having.

It turns out that in our install the native SNMP daemon only likes to run on port 161. Once we swapped ports so that the AIX SNMP daemon runs on 161 and the Net-SNMP daemon runs on 1610 we were 90% of the way there. The rest of the issue was some minor Cisco ACL's and an AIX route command. This is because the host to be monitored is on a diffrent network then the monitoring host.

Thanks for all the help. I plan on getting a propper write up going later on this week to fully explain this approach in hopes that it proves useful.

Looking forward to it. Write to me and I'll work out how you can become a rootvg author and get your signature on it.

And, while I mention it - anyone who has something useful, or even "a long and difficult road" as this may have been -- anyone who is interested in submitting content - I am always looking for contributers! That is one of the great features of a CMS system like Joomla!

Michael