Trigger boot en shutdown of cluster node from loadbalancer

Hi everybody,

We are running a Apache cluster with a couple of dozen nodes behind 2
Ultramonkey loadbalancers (master - slave). The load on the cluster
peaks around 7pm but drops very quickly after 10pm.  To save money on
our electricity bill we currently shut 75% of the nodes down around 11pm
and wake them up again the day after around 9am.

To prevent overload on the cluster when unexpected loads occcur a better
solution might be to monitor the active connections to each of the
nodes.  When the number drops below a certain predefined thresshold some
of the nodes should automatically put to sleep.  When the active
connections exceed another thresshold the nodes should be woken up.

I'm not shure my new idea is good practice.  There is ofcouse the
possibility that the nodes get in an endless 'shut down - boot' loop
when the thressholds are badly choosen.  Has anybody done something like
this before?  Any input is very appreciated.

Regards,
Jochen Billen



_______________________________________________
Ultramonkey-users mailing list
[email protected]
http://lists.vergenet.net/listinfo/ultramonkey-users