Description of Problem: I have several processes working hard, but "top -i" shows nothing. Indeed, just plain "top" seems to be very inaccurate (e.g. giving 99.9% CPU to kswapd) compared to KPM (KDE System Guard 1.2.0) or system-monitor-2.0.0. Version-Release number of selected component (if applicable): procps-2.0.7-21 How Reproducible: Very Steps to Reproduce: 1. start some processes that work hard (e.g. dump -j) 2. start "top -i" 3. compare with e.g. system-monitor-2.0.0 Additional Information: dual pentium
I tried to attach a screenshot, but it failed. Let me know if you would like me to try again later.
-i removes processes that are sleeping or zombie. dump is likely to block on disk i/o, and is therefore considered idle. Look at it in non -i mode and you'll see that it's in "S" state. Exactly how is it inaccurate? We have a few kernel bugs that affect top and ps output.
Created attachment 69626 [details] top vs. system-monitor
Bugzilla finally let me attach a screenshot. I have a "dump -j" running between two local disks. Even with the disk I/O, the compression is chewing up quite a bit of CPU time. system-monitor (All Processes) shows what I would expect to see - several dump processes consuming plenty of CPU. But top (not -i) doesn't show a single dump process, even though it knows the CPUs are only a little bit idle. I'll attach another screenshot shortly: system-monitor (Active Processes) vs "top -i". The only process the "top -i" shows is X, while system-monitor still shows several dump processes. Also, here is the relevant ps output: [kurquhart@oscar kurquhart]$ ps -le | grep dump 100 S 0 11777 11410 0 75 0 - 329 oprof_ pts/5 00:00:01 dump 040 S 0 11778 11777 0 75 0 - 349 oprof_ pts/5 00:00:09 dump 040 R 0 11779 11778 57 85 0 - 345 oprof_ pts/5 00:09:56 dump 040 R 0 11780 11778 57 85 0 - 345 oprof_ pts/5 00:09:56 dump 040 R 0 11781 11778 57 85 0 - 802 oprof_ pts/5 00:09:55 dump
Created attachment 69627 [details] "top -i" vs system-monitor (Active Processes)
Are you still having this problem or has it been fixed in later updates?
Problem is still present in procps-2.0.7-25, which is the latest for RedHat 8.0. If a later version is backwards compatible with RH8.0, I willbe happy to try it; just let me know which version to try.
Your best option is to grab the src rpm for 3.1.15 and rmpbuild --rebuild it and then install.
This problem is fixed in procps-3.1.15-3.i386.rpm. Thanks!