+++ This bug was initially created as a clone of Bug #1890667 +++ +++ This bug was initially created as a clone of Bug #1890361 +++ Recently the s390x 4.5 RHCOS pipeline started failing because the ostree version of fedora32 which is used in the cosa container had the fix for https://bugzilla.redhat.com/show_bug.cgi?id=1867601 Because of this - the root file system is immutable - which is good, but the ostree version in RHCOS 4.5 is an older version which does not have the fix for the above BZ. This causes rollbacks and upgrades to fail. For now, the coreos-assembler is patched to remove the immutable bit, but it would be good if ostree is bumped with the fix for RHCOS 4.5. --- Additional comment from Prashanth Sundararaman on 2020-10-22 01:30:18 UTC --- This is the said fix that would need to be backported: https://github.com/ostreedev/ostree/pull/2179 --- Additional comment from Micah Abbott on 2020-10-22 16:54:49 UTC --- Targeting 4.7 (though the fix should already be present there and in 4.6); we will probably need to clone this for 4.5.z
Higher priority items prevented work from happening on this issue; labeling UpcomingSprint
Do we really need to fix this in 4.5.z? Did we ship 4.5.x s390x bootimages with the immutable bit on or not? I think the way it should work is: - bootimage is 4.5.x, immutable bit is not set - upgrade to 4.6.1 happens using *old* ostree, new deployment still doesn't have immutable bit - reboot - later upgrade to 4.6.2 happens using new ostree, new deployment has immutable bit - old 4.5.x deployment without immutable bit is GC'd - From now on the immutable bit is set correctly Am I missing something?
The bootimage for s390x today does not have the immutable bit set. I guess we don't really need the fix in 4.5.Z. A 4.6.1 -> 4.6.2 upgrade should have this set as outlined above.
OK closing, but if we have reason to believe differently in the future we can reopen.