On 08/17/2018 01:57 PM, Dusty Mabe wrote: > > > On 08/17/2018 01:20 PM, Laura Abbott wrote: >> On 08/17/2018 08:55 AM, Dusty Mabe wrote: >>> I'd love to make logging during early boot not require everyone >>> to happen to get the `console=` entries ordered correctly on the >>> kernel command line. Inevitably we always get it wrong or the ordering >>> of the `console=` lines on a pre-baked works great for some people >>> but not for others. What if there was a way to get those early startup >>> systemd unit messages on all `console=` entries? Maybe there is. >>> >>> I've got some code that I think could help with this. We basically configure >>> the journal in the initramfs to log all messages at 'info' level and above >>> to kmsg. We also have to make the kernel not ratelimit messages during this >>> time. Check out the proposed change and see what you think: >>> >>> https://github.com/dracutdevs/dracut/pull/445 >> >> Probably obvious question but can you clarify what the exact problem >> is with logging in initramfs right now? It sounds like the issue >> is that if something goes wrong early it doesn't get logged to >> the journal so you want to output it to the console? > > Right now journal output will go to the primary console of the machine, > which is the last `console=` entry on the kernel command line. It's been > like this forever. It's expected. > > The problem is that often enough a user either doesn't know which device > is the primary console or somehow doesn't have access to the output of the > device that is associated with the primary console, thus they must play a > guessing game. If somehow they don't have access to that device then they are > out of luck. > > I'm proposing that we log journal output to kmsg during the initramfs (after > switch root that won't be the case because the journald.conf settings from > the root filesystem would take effect), which will mean that output goes to > all `console=` devices and not just the primary one. > > The reason for this request is because we are starting to use ignition [1] > a lot more and debugging troubles in the initramfs is much easier if you don't > have to hunt for the right output device where the magical messages are. Of > course in Fedora CoreOS we could just apply these settings just for CoreOS, > but this is a problem I've had myself for a long time and I'd actually like > to see it applied to all of Fedora. > > Dusty > > [1] https://github.com/coreos/ignition/ bumping this.. i would like some input if anyone has any, whether it be good or bad. Dusty _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx/message/MFMMYGRQ6BCXD2VSV2M7A46ETLB4ODT4/
