On 2/26/19 2:40 AM, Filippo Giunchedi wrote:
Hi all,

On Mon, Feb 25, 2019 at 7:33 PM Andrew Bogott <abogott@wikimedia.org> wrote:
On 2/25/19 12:06 PM, Filippo Giunchedi wrote:

On Mon, Feb 25, 2019 at 6:40 PM Giovanni Tirloni <gtirloni@wikimedia.org> wrote:

  If we poke roles in the firewall so Icinga can reach the VMs and we
define the monitoring::service stuff in Puppet, is that all we need to
shutdown Shinken? Do you think there would be any concerns with going
that route?

I'm not sure, what checks does shinken do ATM that we'd need to port over to icinga? Right off the bat it seems simpler to me to fire up a icinga stretch VM and move over the checks and their puppetization to that VM instead, what do you think?

One of the reasons we haven't done this already is that WMF production monitoring seems to be in constant flux -- it's disappointing to adopt a monitoring tool only to have it deprecated a month later in production.  Is the current version of icinga that's running in prod actually your desired final state there, or is something else up and coming?

Yes I can confirm that the current icinga+stretch combo in production is what we'll stay on. I think there's consensus that icinga is showing its age, however the project of replacing icinga (or even deciding not to!) AFAIK isn't formally on the radar, IOW production icinga deprecation (if any) is many quarters out.

Ok!  So, in that case, icinga-on-a-VM seems like the obvious move to me.  Giovanni, does that suit you OK?