Jerry Jelinek wrote:
Brian Kolaci wrote:
Thats how I came to the conclusion with the current implementation of
getting load averages coming from processor sets rather than the load
running within a zone prohibits the sole use of FSS to consolidate
sendmail servers. The only feasible solution is to use pools/psrsets
and make sure each zone is allocated a proper amount of CPU. That way
its load average algorithm will continue to work.
Although it is not integrated yet, cpu-caps should be available soon.
One idea might be to use a combination of FSS and cpu-caps to really
over-provision a system running lots of consolidated sendmail servers
since the cpu-cap would ensure that any specific zone did not use
too much of the system even though cycles were available. FSS would
ensure that each zone got its share if all of the zones got busy.
I like the idea of cpu-caps, however hard caps are similar to processor sets.
I've requested that caps have two thresholds, one where an alert is sent
and another that stops scheduling anything in that pool. The idea of
the alert is to let capacity planning folks know more resources are needed
without actually holding back resources.
But then that defeats the purpose of the sendmail throttling based on load
averages, or at least doesn't help. The load averages would need to take
into account the caps rather than the processor sets.
zones-discuss mailing list