The following has be reported by IBM LTC: OnDemand Timer patch breaks TOP Hardware Environment: zSeries Software Environment: RHEL3 U2 Steps to Reproduce: 1.deactive OnDemand Timer (echo "0" < /proc/sys/kernel/hz_timer 2.run Top 3. Actual Results: TOP produces error message: INVALID CPU time data! Snap shot of TOP output: CPU states: cpu user nice system irq softirq iowait idle total 100.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% B: PPIcpu00 - INVALID CPU time data! cpu01 100.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Expected Results: Snapshot of good top output with OnDemand Timer active (/proc/sys/kernel/hz_timer = 1) CPU states: cpu user nice system irq softirq iowait idle total 0.0% 0.0% 0.1% 0.0% 0.0% 0.1% 99.7% cpu00 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 100.0% Additional Information:
verified issue on internal test systems ...
This was a known issue when the On-Demand timer patch was implemented, which is why the default is off. However, I'm reassigning this to the maintainer of the procps package, which is what includes the "top" command, to see if something sensible can be done in the case where the On-Demand timer is enabled.
Can we see if procps-3.2.0 fixes this problem?
We could, if you've built it. Do you want me to try FC2 package? [zaitcev@devserv zaitcev]$ ls /mnt/redhat/beehive/comps/dist/3.0E*/procps /mnt/redhat/beehive/comps/dist/3.0E-errata-candidate/procps: 2.0.13-9.3E /mnt/redhat/beehive/comps/dist/3.0E/procps: 2.0.13-9.2E [zaitcev@devserv zaitcev]$ ls /mnt/redhat/beehive/comps/dist/fc2*/procps 3.1.15-5 3.2.0-1 3.2.0-1.1 [zaitcev@devserv zaitcev]$
Ok I built it in 3.0E-errata-candidate Dan
Tried 3.2.1-3. Good news - no division by zero (instant fault on s390), no error messages. Bad news - it just zeroes all columns when counts are not advanced. top - 12:18:27 up 19 days, 3:11, 1 user, load average: 0.03, 0.02, 0.00 Tasks: 43 total, 1 running, 42 sleeping, 0 stopped, 0 zombie Cpu0 : 0.0% user, 100.0% system, 0.0% nice, 0.0% idle Cpu1 : 0.0% user, 0.0% system, 0.0% nice, 0.0% idle Cpu2 : 0.0% user, 0.0% system, 0.0% nice, 0.0% idle Cpu3 : 0.0% user, 100.0% system, 0.0% nice, 0.0% idle Personally, I can live with it. Bill?
Pete, I don't have a problem with it. I'll ask our development team to take a look and comment.
Here are Martin's comments. Martin Schwidefsky <schwidefsky.com> 04/07/2004 12:13 PM To William Goodrich/Poughkeepsie/IBM@IBMUS cc Volker Tosta <TOS.com> Subject Re: Bug 119072 - LTC7041-OnDemand Timer patch breaks TOP I personally can live with the behaviour of procps 3.2.1 to show 0% idle time instead of 100%. It would be preferrable to have a procps that is smart enough to recognize that the system didn't tick in the last 5 seconds but as long as top & ps won't crash anymore I am ok with it. blue skies, Martin Linux/390 Design & Development, IBM Deutschland Entwicklung GmbH Schönaicherstr. 220, D-71032 Böblingen, Telefon: 49 - (0)7031 - 16-2247 E-Mail: schwidefsky.com