Change or disable global status errors right after a director deployment

[slalomsk8er]

Is your feature request related to a problem? Please describe.

Every director deployment results in messages like this: Which make Icinga look bad in the eyes of the users.

Describe the solution you'd like

Make the code check if the director just deployed and based on this increase the timeout or change the message text and color.

Describe alternatives you've considered

Globally increase the timeout.

Additional context

Add any other context or screenshots about the feature request here.

[nilmerg]

Hey, how exactly is the deployment performed?

[slalomsk8er]

We deploy manually ATM.

[nilmerg]

And what does manually mean? Exactly? :wink:

[slalomsk8er]

Clicking on one of the "Ausrollen" links that are distributed all over the director. 😉

[nilmerg]

I cannot reproduce this (with a sleep(120) in my director config).

Any idea what could cause this? @yhabteab

[yhabteab]

I cannot reproduce this (with a sleep(120) in my director config).

As we discussed last time, a director deployment should never prevent Icinga DB from updating the icingadb_instance table, but looking at the Icinga DB web code, I see two reasons why this might happen:

• It's unlikely in this specific situation, if there's no database entries at all in the icingadb_instance table.
• If the instance heartbeat from the database is less than now() - 60. Though, when a Icinga Director deployment would interfere with this in any way, Icinga DB Web would actually render Redis is outdated. Make sure Icinga 2 is running and connected to Redis. instead, but this problem could be caused by the same reasons as in https://github.com/Icinga/icingadb/issues/787 since @slalomsk8er is affected by that issue.

[nilmerg]

If there are multiple rows in icingadb_instance, Icinga DB Web makes sure that the newest (heartbeat desc) is evaluated. So it shouldn't be affected by this :thinking:

[yhabteab]

If there are multiple rows in icingadb_instance

That is not the problem in the referenced issue! The problem is that the active Icinga DB instance inserts a outdated heartbeat instance into the icingadb_instance table while remaining HA responsible, and the passive instance reads this outdated heartbeat like Icinga DB Web does and thinks that the other instance is gone and has to take over HA responsibility, resulting in both instances becoming responsible.

[nilmerg]

thinks that the other instance is gone

and doesn't insert a row in icingadb_instance because of this?

If so, it may be the same reason. But why is this related to a director deployment? (@slalomsk8er wrote above the message (competition) is caused by this every time)