Icinga 2 v2.9 introduced performance related changes inside the configuration compilation and activation order. This was to ensure a) no unwanted notifications b) use available CPU resources to speed up the overall validation process. These changes had a bad effect on configuration depending on a specific activation order, and slowed it down with many config objects of a specific type. The Icinga Director depends on get_host() being called in service objects to support specific service set overrides. In case you’re having trouble here, v2.9.2 is for you.
In addition to these fixes, v2.9.2 doesn’t crash on Fedora 28 and API queries anymore. We’ve also fixed a regression with MatchAny in match/regex functions inside the DSL. Special thanks this time to Christopher who has fixed the cluster-zone health check.
Users have reported that custom attributes are not updated inside the IDO database after upgrading from v2.8. Unfortunately we were not able to reproduce this after many hours, so the workaround is to purge all config/status tables and restart Icinga 2. This populates the data again.
Check the Changelog for more details, official packages have been published already.
We’re also testing a patch which avoids disconnecting many parallel CSR signing requests. This includes a user config option scheduled for v2.10 and follows more features like namespaces, improved acknowledgement notification handling, scheduled downtime options and more.
There’s also a long lasting investigation on TLS timeouts with (re)connects in #6517. Finally, a patch has been merged and needs testers in large scale environments, snapshot packages are available on packages.icinga.com. With your test feedback, we hope to release v2.10 in CW41. Thanks in advance!