mjhjmtu

I installed trace-cmd.x86 package on CentOS 7 because I was planning to trace some latency issue in my application.

But then, I noticed once awhile, the boot up time took like 2+ mins.

Then, I looked into /var/log/messages, and I found:

Oct 22 08:57:13 heilinux systemd: Mounted /boot.
Oct 22 08:58:13 heilinux systemd-udevd: slow: '/usr/bin/systemctl is-active trace-cmd.service' [739]
Oct 22 08:58:13 heilinux systemd-udevd: worker [559] /module/mei is taking a long time
Oct 22 08:59:12 heilinux systemd: systemd-udev-settle.service: main process exited, code=exited, status=1/FAILURE
Oct 22 08:59:12 heilinux systemd: Failed to start udev Wait for Complete Device Initialization.
Oct 22 08:59:12 heilinux systemd: Unit systemd-udev-settle.service entered failed state.

For some reason, systemctl took like a minute to figure whether trace-cmd is active (although system-analyze blame indicates systemd-udev-settle.service took 2+ mins. I believe systemd-udev-settle.service was using systemctl to check whether trace-cmd was active, taking a minute).

And then, systemd-udev-settle.service was marked as failed.

Any idea what's going on?

The delay doesn't happen on every single reboot but approximately 50% of the time.

One solution is to erase trace-cmd package, which I would like to keep to trace latency in my application.