Hi all, I figured it might be worth chiming in with an example from users (even though we're not quite novice users anymore). On the one hand we really like cephadm - it's beautiful with central reporting of versions, status, default ability to perform rolling updates, placing services, etc. A minor drawback is that we can no longer easily handle things in salt, but I think that's a matter of having to accept that we can't both eat the cake and have it (since both are different management tools). However, we recently had another snafu - suddenly cephadm upgrade doesn't do anything - it just waits. I'm sure this is a trivial bug/config setting somewhere, but there are pretty much no log messages about, and what would have been trivial with standard packages (just force an upgrade, one node at a time, if necessary) suddenly becomes a bug-hunt that might take a day or two. This is most definitely not an argument against Cephadm. I think it's a great solution we'll stick with, and I'm sure it will only get more stable. Still, I think the challenge for non-expert users is that *default* deployments and maintenance can be easier with container solutions, but when you need to step out of that and debug/customize things, the additional layer(s) create a barrier. Cheers, Erik _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |