Too many requests after server goes down and up

[limo]

Hi all,

I do not know if somebody has same feeling like me but I think that there is some problem in poller.c or somewhere. It is not realy bug, maybe feature request, and it belongs to queue management.

I thing there should be some check befor poller tries to catch some data. That check should do this:

- isn't it too late to catch data ? (we do not need catch value 5 minutes later, if we have next values each minute). Sometimes when I see log, I think, that if something is unreachable some time, poller tries to get same values more times in same time. Am I wrong?

- Maximum host requests per time check

- Maximum host group requests per time check

Maybe in future, but I know this will be harder to achieve:

- Maximum bandwidth per host

- Maximum bandwidth per group

There is question, what to do with queries which are over quota. Simplest is to drop them, of course...

Is this scenario/concept acceptable for user/developers?
Thanx.
Lukas

[Alexei]

Why do you need this functionality? Something related to slow network links?

[limo]

Yes, sometimes group of servers or server can be behind slow expensive link. We need real data on link too, not only monitoring :) In today situation, zabbix can do too much traffic on that line without our control. It can happen due to problem which I discused here.

It is smaller problem to loose some monitored data than make line unusable. Maybe I am wrong and users does not need this. But I think if this will be implemented, more problems will be solved.

- problem with more same checks in same time

- problem with burst of checks (checks will be spreaded in time)

- problem with slow lines