exclude or skip wildcard match for poller and discovery
Hello,
We are currently running Pro r8993 with 650+ devices and will be adding another 400 or so shortly, to support this we will try a deployment as described in some detail by Markus: http://blog.best-practice.se/2017/10/scaling-observium-horizontally.html
We do however also have another edge case to deal with in the form of an infrastructure that we cannot look into but can send out of, the inhibitor is political and not technical.
We are halfway to solving it technically however by using a poller remotely with wildcard host name matching as the 'remote' stuff has a common domain part.
Our other 'normal' pollers however need to be told to exclude or skip polling / discovery of the remote devices.
Can a wildcard skip or exclude be added to the poller-wrapper script for this purpose?
Picture attached for reference.
Regards
[cid:image001.jpg@01CE9215.D39244B0]https://www.it.ox.ac.uk/
Richard Wigzell Technical Lead, Storage Platforms IT Services | University of Oxford 13 Banbury Road, Oxford OX2 6NN
T: 01865 616282 www.it.ox.ac.ukhttps://www.it.ox.ac.uk/
I was aware of this use case when I wrote the guide and made a patch that solves this case. Basically I modified the poller-wrapper.py to accept group-ids and poll a specific group only. You can then create a host-group in observium for the remote poller to poll and create another group with "all other hosts" and have poller 1 and poller 2 poll that group instead. Just make sure that all your hosts are in some of these two groups to have them all polled.
The patch was not accepted however as the main plan is to register pollers in the database and be able to control this via Observium in a more controlled manner. If you want to try it either way you can find the patch here: http://jira.observium.org/browse/OBS-2245
/Markus
Hello, thanks for this.
I think that we could probably make that work for us in lieu of DB registered pollers.
Two questions: will the patch work for the current poller-wrapper? And how far along are registered pollers in the database? i.e. is this something that’s likely to come along shortly? If not I think we will try a deployment with the patch.
Regards
[cid:image001.jpg@01CE9215.D39244B0]https://www.it.ox.ac.uk/
Richard Wigzell Technical Lead, Storage Platforms IT Services | University of Oxford 13 Banbury Road, Oxford OX2 6NN
T: 01865 616282 www.it.ox.ac.ukhttps://www.it.ox.ac.uk/
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Markus Klock Sent: 09 January 2018 13:53 To: Observium observium@observium.org Subject: Re: [Observium] exclude or skip wildcard match for poller and discovery
I was aware of this use case when I wrote the guide and made a patch that solves this case. Basically I modified the poller-wrapper.py to accept group-ids and poll a specific group only. You can then create a host-group in observium for the remote poller to poll and create another group with "all other hosts" and have poller 1 and poller 2 poll that group instead. Just make sure that all your hosts are in some of these two groups to have them all polled.
The patch was not accepted however as the main plan is to register pollers in the database and be able to control this via Observium in a more controlled manner. If you want to try it either way you can find the patch here: http://jira.observium.org/browse/OBS-2245
/Markus
Man, do we need to just accept this swepatch? :D
It's probably easier to do it with wildcards or something. I'll look at that patch again.
Adam.
Sent from BlueMail
On 9 Jan 2018, 13:53, at 13:53, Markus Klock markus@best-practice.se wrote:
I was aware of this use case when I wrote the guide and made a patch that solves this case. Basically I modified the poller-wrapper.py to accept group-ids and poll a specific group only. You can then create a host-group in observium for the remote poller to poll and create another group with "all other hosts" and have poller 1 and poller 2 poll that group instead. Just make sure that all your hosts are in some of these two groups to have them all polled.
The patch was not accepted however as the main plan is to register pollers in the database and be able to control this via Observium in a more controlled manner. If you want to try it either way you can find the patch here: http://jira.observium.org/browse/OBS-2245
/Markus
observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
participants (3)
-
Adam Armstrong -
Markus Klock -
Richard Wigzell