A new RC1 of MAAS 3.4 was released on August 2nd 2023, which seems to break DNS. New records added to the MAAS DNS are not resolvable through queries until the MAAS snap is restarted. This is breaking deployments.
@billwear @craig-bender is anything known about this issue?
HI @kevin-reeuwijk, can you please file a bug about this, attaching regiond logs?
@kevin-reeuwijk, are you using this in production?
@billwear 3.4 RC1 is deployed at several customers and internally, since we need the bugfix it contains for successful disk erasure of nodes deployed with custom images. One of those customers is in production, but fortunately not immediately affected as they are not deploying new clusters at the moment. Our internal environments and customers with lab build are however directly affected.
3.4 RC1 has been working great so far, but the newer build of RC1 that got deployed August 1st/2nd has broken things. If the previous RC1 build was made available again, we could roll back to it.
The previous RC1 build (as well as virtually all 3.4 builds) have had sporadic issues with DNS seizing to take new changes (after MAAS was running for weeks/months), but this was always fixable with a snap restart. Now though, the issue is persistent.
was going to share the usual words of caution about deploying an RC, but i see you know what that means and weighed the risks carefully. tough call, nice work limiting exposure.
@kevin-reeuwijk thanks again for the bug report!
For beta and edge channels I’d suggest using snap refresh --hold=forever just not to break all environments at once.
This topic was automatically closed after 40 hours. New replies are no longer allowed.
