loading balance

[proteinx]

Hi, i am a new system admin for a HP XC cluster, just wondering whether there is fee utility provided from HP to monitor real time loading balance and CPU/mem usage for each working node and head node.

And, how people here judge whether a head node is overloaded or not? by CPU usage or by other things? I mean, I have to make a judgement and sometime might terminate people's jobs running on the cluster..

[Steven E. Protter]

Shalom,

top

Is sufficient for what you wish. Run it on both nodes.

It can also be run with output directed to a file, looking for problems and taking actions on that basis.

SEP

[proteinx]

but if I have 128 nodes, I have to type 128 top on each node?

I guess there are must be a better way, something like real time chart keep recording what is going on each node. Once something wrong or overloaded, an alert will be generated and reported...

It must have such software, not that difficult, I can make one for myself if I had enough time for this.

[Vitaly Karasik_1]

I'm not familiar with HP XC cluster, but there are several products for monitoring clusters with many nodes. See Ganglia (http://ganglia.sourceforge.net/), for example.