[2019-03-16 16:29] Harald Dunkel <ha...@afaics.de> > I am running ext4 instead of reiserfs today, but logging fsck has still > a *severe* impact on boot time performance. We have a few Debian file > servers in the office, e.g. providing /home/* via NFS. They are managed > remotely using some serial-over-line technology instead of a vga console. > > A few months ago such a server went down without a clean umount. On the > next boot it was busy for >2h to show billions of lines about some tiny > repairs it has performed. Mo end was in sight. Then I gave up, interrupted > fsck, booted the host in single user mode, and rerun fsck writing to a log > file instead of the serial line. It was completed within 15 minutes. No > serious problems.
I see. How fine grained control you need? Do you need separate controls to enable/disable logsave for `checkroot.sh' and `checkfs.sh'? I consider two possible implementations for current feature request -- either make FSCK_LOGFILE variable configurable via /etc/default/{checkfs,checkroot} or add variable `WANT_LOGSAVE' there. Not sure, which is better. Opinions? Just in case. Dear submitter, you are aware, that init.d scripts are conffiles, aren't you? -- Note, that I send and fetch email in batch, once every 24 hours. If matter is urgent, try https://t.me/kaction --