my Thinkpad X230 suddenly started booting slow, a few weeks ago. It was booting in about 20s, now it takes about 120s. I can't find an obvious reason. It seems, it is hanging after the filesystem checks. The hard disk I/O led is not flashing, so it's probably not a slow I/O issue.
No systemd service or mount takes unusually much time to startup.
$ systemd-analyze blame | head -5 9220ms firstname.lastname@example.org 783ms email@example.com 387ms laptop-mode.service 236ms systemd-binfmt.service 150ms lvmetad.service
The startup time for the kernel seems pretty usual, about 100s for the userspace is definitely unusual for my configuration.
$ systemd-analyze time Startup finished in 14184ms (kernel) + 99418ms (userspace) = 113603ms
The boot graph (systemd-analyze plot, http://henning.orgizm.net/arch/boot.svg) indicates, the hanging (about 88s) happens between "mnt-platter.mount" and "local-fs.target". The filesystem for /mnt/platter is ext4, contains no errors and mounting in a live system is instantaneous.
I tried booting with an initcpio generated by the last but one version of mkinitcpio (0.12.0-3) and it makes no difference.
Does someone of you have the same problem or ideas for further debugging it?
Hum, well I have found that sometimes the parallelization of systemd boot-up can make it seem like one thing is causing a problem when really it is something else.
One sure fire way to figure out what it is, by had go through your journal and manually look through all the systemd messages from boot. Look for the timestamp it says it is Starting a service and subtract the timestamp when it says it has Started the service.
By doing that I figured out that a problem I was having was tor.service was configured in such a way that at shutdown the Tor server would wait a long time before finishing taking down the tor server.
Examining "journalctl -xb" brought up a timeout waiting for a block device. The problem was an erroneous entry in /etc/crypttab.
Thanks for your help! I should have thought of this before posting...