On Wed, Oct 17, 2012 at 11:04:28AM -0400, Simo Sorce wrote: > > 6. Strong QA exploring corner cases and data corruption (procedure tbd) > > 7. Clear, simple, and Fedora-centric disaster recovery documentation > I am not sure 6/7 are strong requisites. We certainly shipped stuff that > was hard to deal with when it broke before. > I would love to see this done for F19 but I think it would be unfair to > block the feature just on point 6/7 unless there is a significant higher > risk of issues with the journal than there ever was with malfunctioning > (r)syslog. > > Is your worry that because the journal file is a binary format and not a > plain text file that corruption of the journal files is more likely / > more problematic ? It isn't that systemd/journald are particularly unreliable or risky inherently. It's just a big change, which is a big risk *for Fedora*, and we should be extra diligent so that the reception is positive. It doesn't take many undiscovered corner cases to make the change a fiasco. The quality assurance process here doesn't need to be turned up to 11, but it does need to at least be thorough. Likewise the DR documentation will help ease the transition simply by existing. (And here, I'm not just talking about recovering from systemd failures, but, for example hardware failure that corrupts the journal, or simply takes down the machine so dead-system forensics are required.) > Is your worry that because the journal file is a binary format and not a > plain text file that corruption of the journal files is more likely / > more problematic ? I don't know if it's more likely. It's seems to be almost a tautology that corrupt binary formats are more problematic. Just the other day I hit something that caused a corrupted journal and made journalctl stop telling me things. (bz 865091) I really was at a loss as to what to do. That might have just been a weird glitch, but... well, isn't that the point? > I was wondering if /var/log/messages could be turned into a named pipe > to which journal start spitting out stuff in syslog format when someone > tails it ? I think the semantics of named pipes are enough different that this might actually make the transition worse, as nice as it would be for something like that to magically work. -- Matthew Miller ☁☁☁ Fedora Cloud Architect ☁☁☁ <mattdm@xxxxxxxxxxxxxxxxx> -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel
