Description of problem: Short: Journal lacks roughly 12 hours of kernel messages in an otherwise readable journal log. Long: Soon after a 'dnf install kernel' dmesg reports corruption in systemd journal file and that system.journal is being replaced by a new one. journalctl --verify shows no problems; btrfs can read all files (to /dev/null) without triggering any fs checksum mismatch errors. Upon reboot, journalctl --verify lists journal corruption; still btrfs reads files to /dev/null no errors. But then I also can't find the corruption message I saw with 'dmesg | grep -i corrupt' in the previous boot. And then I see 'journal -k -b -1' shows no kernel messages for almost 12 hours, except some audit messages. For sure there were a bunch of kernel messages today, but none of them are locateable in the journal. Could be related to bug 1294002 (same system) but manifests differently Version-Release number of selected component (if applicable): systemd-222-10.fc23.x86_64 How reproducible: unknown, some form of journal corruption is inevitable but I don't know what triggers it and manifestation of corruption is different each time; this is the first time I've noticed specifically kernel messages missing. Steps to Reproduce: 1. Uncertain 2. 3. Actual results: Kernel messages in previous boot appear to be missing. Expected results: Kernel messages for prior boot should be preserved more reliably. Additional info:
Created attachment 1111647 [details] journal file --verify says is corrupt
Created attachment 1111648 [details] journal others These are not marked as corrupt by --verify, but go with the other log file, so this is included just to have a complete submission.
Ultimately this is the same problem as bug 1294002. *** This bug has been marked as a duplicate of bug 1294002 ***