Switch to sqlite rpmdb backend #795
Conversation
|
Let's keep this as draft until the new year and e.g. one more round of releases to be safe. |
There won't be any support for writing to the bdb backend in f34, so e.g. pkglayering won't work (and obviously even composes wouldn't work once we move cosa to f34). All the production streams now have an f33 barrier release (as part of the resolved workarounds), so it should be safe to switch to sqlite now. For the other rpm-ostree-based variants, we'll probably just flip the default in rpm-ostree to sqlite in f34. But at least in FCOS, we can do this now so that we flush out any issues earlier and squash the ugly warnings when querying the rpmdb. Closes: coreos/fedora-coreos-tracker#623
The main risk is if a user tries to do pkglayering on top of an f33/sqlite commit from an f32 deployment. Now, that should be pretty rare because we do have a barrier f33 release, and they'd be skipping over that. But e.g. they could have reverted from the barrier release back to the f32 fallback, and then tried a direct
is to widen the distance from that possibly happening. But then again, I think at this point we really require nodes to go through the barriers, because our maintenance efforts rely on assuming that all the precise state mutations we've rolled out actually took place. |
|
New year and we've had another round of releases, so I think we're ready to do this! |
|
This already had two approvals, so just restarted CI and marked for auto-merge. |
|
This is causing coreos/rpm-ostree#2435 |
|
Possibly some variant of openSUSE/libsolv#386 |
|
OK, let's revert for now: #812. |
There won't be any support for writing to the bdb backend in f34, so
e.g. pkglayering won't work (and obviously even composes wouldn't work
once we move cosa to f34).
All the production streams now have an f33 barrier release (as part of
the resolved workarounds), so it should be safe to switch to sqlite now.
For the other rpm-ostree-based variants, we'll probably just flip the
default in rpm-ostree to sqlite in f34. But at least in FCOS, we can do
this now so that we flush out any issues earlier and squash the ugly
warnings when querying the rpmdb.
Closes: #623