Bug 89959
| Summary: | "sa" shows wrong cpu time for killed processes | ||
|---|---|---|---|
| Product: | Red Hat Enterprise Linux 3 | Reporter: | Need Real Name <vpose> |
| Component: | psacct | Assignee: | Charlie Bennett <ccb> |
| Status: | CLOSED ERRATA | QA Contact: | |
| Severity: | medium | Docs Contact: | |
| Priority: | medium | ||
| Version: | 3.0 | CC: | bernds, botsch, gasi, jan.iven |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | i686 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2004-12-21 03:50:31 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
|
Description
Need Real Name
2003-04-30 11:33:36 UTC
We are having the same problem, but it is worse than simply killed processes. Kernel 2.4.18-10 is fine. Upgrading to 2.4.18-27.7.x has broken processing accounting. Here is an example of it being broken: > time openssl speed (SNIP) 314.796u 0.056s 5:21.43 97.9% 0+0k 0+0io 232pf+0w > lastcomm |grep openssl openssl dwb7 stdin 1611.81 secs Sun May 4 19:47 as you can see, openssl speed took about 5 minutes to run. But lastcom shows the time being about 5 times that number. Reproducible 100% of the time on ANY computer running 2.4.18-27. This is a HUGE problem as it results in users being overcharged for cpu time. We believe we've found the problem.
In 2.4.18-27, the HZ value is set to 512 instead of 100 for i686 and athlon (but
for some odd reason, still set to 100 for i386 and i586). You can confirm this
by grepping for "_HZ" in the configs for kernel source:
> grep "_HZ" *
kernel-2.4.18-athlon.config:CONFIG_HZ=512
kernel-2.4.18-athlon.config:CONFIG_HZ=512
kernel-2.4.18-athlon-smp.config:CONFIG_HZ=512
kernel-2.4.18-athlon-smp.config:CONFIG_HZ=512
kernel-2.4.18-i386-BOOT.config:CONFIG_HZ=100
kernel-2.4.18-i386.config:CONFIG_HZ=100
kernel-2.4.18-i586.config:CONFIG_HZ=100
kernel-2.4.18-i586-smp.config:CONFIG_HZ=100
kernel-2.4.18-i686-bigmem.config:CONFIG_HZ=512
kernel-2.4.18-i686-bigmem.config:CONFIG_HZ=512
kernel-2.4.18-i686.config:CONFIG_HZ=512
kernel-2.4.18-i686.config:CONFIG_HZ=512
kernel-2.4.18-i686-debug.config:CONFIG_HZ=512
kernel-2.4.18-i686-debug.config:CONFIG_HZ=512
kernel-2.4.18-i686-smp.config:CONFIG_HZ=512
kernel-2.4.18-i686-smp.config:CONFIG_HZ=512
For things like proc file output, a conversion is done so that you see the right
values in realtime:
from linux-2.4.18-variablehz.patch:
+ hz_to_std(task->times.tms_utime),
+ hz_to_std(task->times.tms_stime),
+ hz_to_std(task->times.tms_cutime),
+ hz_to_std(task->times.tms_cstime),
but, if you look at acct.c, which records the accounting files, this conversion
is not done.
So, either acct.c needs to be patched to do the conversion, or you have to go
download later source for gnuacct which supports the --ahz option allowing you
to specify the hz value of 512 to get correct output. Or, you have to edit the
configs to define CONFIG_HZ as 100 for the appropriate archs instead of as 512.
*** Bug 114617 has been marked as a duplicate of this bug. *** *** Bug 117817 has been marked as a duplicate of this bug. *** This is fixed in FC3t3 and AS2.1. Moving this bug to RHEL3 so that it gets fixed there, too. Erratum RHBA-2004:497-05 An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2004-497.html |