The technique used by proxies to accumulate data while server is down is a very valuable one indeed. But well I think it'll be great if newly arrived data would have a somewhat higher priority. Otherwise all the nodata() triggers flash until all the data is consumed by zabbix.
-
-
anyone got same problem?
as far as I know developers added a special check so that zabbix server would not flash nodata triggers if it's up from being down for a long time. It'll be nice to have smth like that for proxies setup, because even after being 5 minutes down triggers with nodata flash, as they hadn't consumed all the data from proxies yet. -
Newer data cannot be processed before old data. This would break order of the event generation causing all sorts of problems.
You may consider configuring of host availability triggers so that they would depend on proxy availability. This may resolve your issue.Comment
-
well actually it wouldn't. The proxy is avaliable and the host it monitors is up and running, but there was no data about that host for the time server was down(say 20 minutes). So it waits for several minutes(say we have nodata(120) and after that server checks to see that last time he got info about that item was more than 2 minutes away so the trigger changes state to PROBLEM.
Maybe there should be some indicator, that analyzed how far behind proxy servers current state is. (inspired by mysql's seconds_behind_master) So that this indicator would also be taken into consideration by nodata() function?
I'm thinking about not useing nodata() in triggers right now.Comment
-
I have exactly the same problem. How could it be fixed?Comment
-
you could try adding a condition on proxy lastaccess parameter. it records unix timestamp when proxy was last seen.
see http://www.zabbix.com/documentation/...nternal_checks for more information
edit : oops, fix item return descriptionComment
Comment