Configuration never fully dispatch
ymolinet opened this issue · 5 comments
hi all,
I have an issue with Shinken (install from the GIT repo) on debian 9 (upgrade from debian 8)
The configuration is check with a OK in result.
After that, Arbiter begun to contact each scheduler but some are offline or unavailable.
In debug, Arbiter try to ping each and some are setting to dead_state (it's normal for me)
but, never the configuration is sent. For example, the broker, installed on the same computer, stay "Waiting for initial configuration" and the scheduler that are online don't received the new configuration.
Some help are welcome to fix this,
Thanks,
Yannick
Just curious; even after applying the changeset from PR #1971? I found out a while ago after installing from Git repo that the broker had a bug where it would never go receive the initial configuration. Applying this patch makes it at least fully start.
hi,
it seems that the configuration is never send to the broker if all schedulers, pollers and reactionners are not alive or in dead state.
But pass each scheduler, each poller and each reactionner in dead state can take severals minutes, and if many schedulers, pollers, reactionners are offline it increased the time ...
Currently, 3 hosts (scheduler + poller + reactionner) are offline, so the dispatch can take more than one 1 hours ...
Could you paste some configuration samples from your shinken services configuration ?
I never had such a behaviour...
My configuration contains a lot of poller, reactionner and scheduler.
The dispatch begin only when the arbiter is successfully connect with each service, so it can pass a long time before the dispatch.
It's maybe a configuration issue ?
Is the dispatch muti process ?
In my configuration, it seems that the arbiter dispatch the configuration realm by realm with only one process...