High CPU utilisation on Xorg when fsck is running
 

Hello,

I'm running into a situation where my system's CPU utilisation jumps from near-zero (1-5% during idle runtime) to 40-60% (I've seen as high as 63%) while fsck is running on my second HD. The odd thing (to me) is that the CPU utilisation as seen by top and system monitor as being tagged to the Xorg process. I'm invoking fsck using 'sudo fsck.ext3 -cf /dev/sdb1', and once it hits the 'Checking for bad blocks (read-only test)' step is where it spikes.

The system is equipped with a P4 3.0GHz Hyper-threaded CPU, a little behind the times for sure. Running with Ubuntu 9.10 in GNOME. I've also caused the same spike in CPU usage by opening an SSH session from my mobile phone, invoking the same command as above, with the same results (further puzzling me, since I can't figure out why Xorg would get involved on a command issued by an SSH user).

Is this near the norm for this sort of command, or is something going a bit awry? I hesitate on letting this rig run with such high CPU usage, seeing as I have trouble getting memtest86 to complete (due to overheat) without starting it from a cold boot; I've seen that a bad block check like this can get rather lengthy.

Sounds like you have a general overheating problem; that's not normal. I'm wondering if during the fsck operation you are generating too much heat and the cpu is throttling down due to excessive heat giving you, in effect, a slower cpu and greater cpu utilization as a result. What kind of temperatures are you getting in your bios setup?

First thing I would try would be opening the case and cleaning out the dust bunnies. Especially, check out your cpu heatsink and fan and blow out any dust you find there.

I blew out the dust, the heat sink was caked about 3/4 inch in gunk; the previous user (my fiance) wasn't that good about cleaning, and until I took her system over it just built up; half a can of compressed air later, it's clean, and that took care of those issues.

It seems that the problem with the fsck was the drive itself; I took a look over some of the logs (if I remember correctly, dmesg had the details, but I've slept since then) and quite a few blocks were getting tagged as bad, which seems to have been my root cause on fsck's runtime; my other system's 2nd HD, swapped into this system, had the same usage but much quicker runtime on the same check.