Monitoring of HeartBeat in MC/SG

[Preeti Khanna]

Hi
Can anybody help me in Monitoring heartbeat of MC Service Guard ? and also i want to Know how a Floating Server Works.

Thanks in advance

[Rumen Ginev]

HP's standard tool for network traffic tracing is nettl. You can also use nettladm -interactive user interface to nettl.

Another good tool is tcpdump. It is free and is available as depot (11.00 and 10.20):

http://hpux.connect.org.uk/hppd/hpux/Networking/Admin/tcpdump-3.4/

To monitor the heartbeat messages you have to filter the output from one of the above tools. You may filter the port used by the heartbeat. It is listed in /etc/services, but usualy is 5300.

Rumen

[Anthony Goonetilleke]

By a floating server I presume you mean a package that has a floating IP within MC/SG
What happens is that you give your two machines two IP addresses that are fixed and then you assign a third IP address to your package. This IP address floats between wheich ever machine the package is running on. If you currently have MCSG look in /etc/cmcluster// and you should find the package config files with the floating IP in it.

What exactly do you want to monitor with the heart beat LAN? You can use nettl or if you simply want to adjust the parameters go to /etc/cmcluster/ and edit the following lines:

HEARTBEAT_INTERVAL xxxxx (microseconds)
NODE_TIMEOUT xxxxx (microseconds)

I am not sure if this answers your question but if you could be more specific I will try and help.

[melvyn burnard]

Please elaborate on what you are looking for in the way of "monitoring" the MC/SG Heartbeat.

A "floating server" is a term often used to refer to the package which is set up using MC/ServiceGuard, as it appears to be a server node out on teh network which "floats" or relocates from system to system in the MC/SG cluster, using the "floating" or relocatable IP address
May I suggest you take a look at :
http://www.docs.hp.com/ha
and review the Using and Managing MC/ServiceGuard

[Emil Velez]

You can also use cmsetlog to change the logging level of MC/ServiceGuard so every time it sends a heartbeat it logs a message to the syslog and every time it receives a message it also logs that.

man cmsetlog will provide you example command lines. The heartbeat is a periodic event.

[Preeti Khanna]

Hi
I am unable to find the man page of cmsetlog . what's wrong?
Also I want to know how user's are logging into these two server . It's depends upon the load basically but i am seeing that whenever user logs it's always goes to primary server i didn't find any user in the Secondary Node. where the user limit is defined in the Clusterning.
But Clusterning is working fine.
2. Where can i find the log of Heartbeat.
Thanks.

[melvyn burnard]

There is no man page for cmsetlog as this is an unsupported command in /usr/contrib/bin
Also, if you turn up logging, you will be filling up syslog.log much faster, thereby increasing disk space useage, and if you make a mistake on the levels and moduiles, it will be a BIG mistake.
There is no real reason you need to monitor the heartbeat as the syustem does this, so why are you wanting to do this?
Also, if your users are NOT logging into the package or floating ip address/hostname, then you need to ask why not.
If they are using the floating ip/hostname, then they will end up logged in to the node that is currently running the package, so again I do not understand your question/problem.

[Preeti Khanna]

Hi Melvyn
Thanks for your reply . Actually i want to write a script whenever cluster fails a mail should come to me indicating that cluster fails .That's why i want to see the log/monitoring of Heartbeat .coz if heartbeat is not healthy then it causes problem in forming cluster.
Can you just please help me in monitoring cluster and whenever it fails mail will reach to me stating that Cluster is not working.

Thanks
Preeti

[melvyn burnard]

I would suggest you look at using the EMS services for this, and also read the manuals available at:
http://www.docs.hp.com/hpux/ha

These will give you lots of information.
If you have not already done so, I would recommend the MC/ServiceGuard training as a good track to follow.