Bug 179487

Please run:

bash -x /usr/share/rhr/external/clock-test.sh &> clock-test.log & sleep 5; kill %1

and attach the output file to this bug.

Created attachment 123935 [details]
Log of clock-test from CORE2 test RHEL3 Upd 4

Please Note: that this was observed on an egenera pblade with RHEL 3 Upd, and
reproduced on a AMD64 whitebox running RHEL3 Upd 3. 

Mike Gahagan requested the info in the included log file.

Here is the output from an egenera pblade running RHEL4 Upd 2:

more clock-test.log
+ LIMIT=600
++ grep '^processor' /proc/cpuinfo
++ wc -l
+ NUM_CPUS=8
+ nanotime='date "+%s+(%N/1000000000)" | bc -l'
+ echo 'Watching the system clock for 600 seconds.'
Watching the system clock for 600 seconds.
++ eval date '"+%s+(%N/1000000000)"' '|' bc -l
+++ date '+%s+(%N/1000000000)'
+++ bc -l
+ cur_time=1138738030.60730400000000000000
+ (( s=0 ))
+ (( s < LIMIT ))
+ sleep 1
+ cpu=0
+ prev_time=1138738030.60730400000000000000
++ /usr/bin/taskset 1 bash -c 'date "+%s+(%N/1000000000)" | bc -l'
+ cur_time=1138738031.61461700000000000000
++ echo 1138738031.61461700000000000000 - 1138738030.60730400000000000000 - 1
++ bc -l
+ jitter=.00731300000000000000
++ echo '(-1.1<.00731300000000000000) && (.00731300000000000000<1.1)'
++ bc -l
+ '[' 1 == 0 ']'
+ (( s++ ))
+ (( s < LIMIT ))
+ sleep 1
+ cpu=1
+ prev_time=1138738031.61461700000000000000
++ /usr/bin/taskset 2 bash -c 'date "+%s+(%N/1000000000)" | bc -l'
+ cur_time=1138738032.62658200000000000000
++ echo 1138738032.62658200000000000000 - 1138738031.61461700000000000000 - 1
++ bc -l
+ jitter=.01196500000000000000
++ echo '(-1.1<.01196500000000000000) && (.01196500000000000000<1.1)'
++ bc -l
+ '[' 1 == 0 ']'
+ (( s++ ))
+ (( s < LIMIT ))
+ sleep 1
+ cpu=2
+ prev_time=1138738032.62658200000000000000
++ /usr/bin/taskset 4 bash -c 'date "+%s+(%N/1000000000)" | bc -l'
+ cur_time=1138738033.63763900000000000000
++ echo 1138738033.63763900000000000000 - 1138738032.6265820

(In reply to comment #2)
> Created an attachment (id=123935) [edit]
> Log of clock-test from CORE2 test RHEL3 Upd 4
> 
> Please Note: that this was observed on an egenera pblade with RHEL 3 Upd 4, and
> reproduced on a AMD64 whitebox running RHEL3 Upd 3. 
> 
> Mike Gahagan requested the info in the included log file.
> 
> Here is the output from an egenera pblade running RHEL4 Upd 2:
> 
> more clock-test.log
> + LIMIT=600
> ++ grep '^processor' /proc/cpuinfo
> ++ wc -l
> + NUM_CPUS=8
> + nanotime='date "+%s+(%N/1000000000)" | bc -l'
> + echo 'Watching the system clock for 600 seconds.'
> Watching the system clock for 600 seconds.
> ++ eval date '"+%s+(%N/1000000000)"' '|' bc -l
> +++ date '+%s+(%N/1000000000)'
> +++ bc -l
> + cur_time=1138738030.60730400000000000000
> + (( s=0 ))
> + (( s < LIMIT ))
> + sleep 1
> + cpu=0
> + prev_time=1138738030.60730400000000000000
> ++ /usr/bin/taskset 1 bash -c 'date "+%s+(%N/1000000000)" | bc -l'
> + cur_time=1138738031.61461700000000000000
> ++ echo 1138738031.61461700000000000000 - 1138738030.60730400000000000000 - 1
> ++ bc -l
> + jitter=.00731300000000000000
> ++ echo '(-1.1<.00731300000000000000) && (.00731300000000000000<1.1)'
> ++ bc -l
> + '[' 1 == 0 ']'
> + (( s++ ))
> + (( s < LIMIT ))
> + sleep 1
> + cpu=1
> + prev_time=1138738031.61461700000000000000
> ++ /usr/bin/taskset 2 bash -c 'date "+%s+(%N/1000000000)" | bc -l'
> + cur_time=1138738032.62658200000000000000
> ++ echo 1138738032.62658200000000000000 - 1138738031.61461700000000000000 - 1
> ++ bc -l
> + jitter=.01196500000000000000
> ++ echo '(-1.1<.01196500000000000000) && (.01196500000000000000<1.1)'
> ++ bc -l
> + '[' 1 == 0 ']'
> + (( s++ ))
> + (( s < LIMIT ))
> + sleep 1
> + cpu=2
> + prev_time=1138738032.62658200000000000000
> ++ /usr/bin/taskset 4 bash -c 'date "+%s+(%N/1000000000)" | bc -l'
> + cur_time=1138738033.63763900000000000000
> ++ echo 1138738033.63763900000000000000 - 1138738032.6265820

Okay, it looks like taskset 1.3.0 (in RHEL3) needs different arguments from
taskset 1.4.0 (in RHEL4). clock-test.sh does not work properly on RHEL3 *but*
still returns success. I will attach a fixed clock-test.sh shortly.

Created attachment 123995 [details]
fixed clock-test.sh

This should work properly on RHEL3 and RHEL4 (tested on i386)

I installed the fixed clock-test.sh on my RHEL3 U4 pblade and it looks like its
fixed. Thanks.

Any estimates when the hwcert kit will be respun to pick this up? We're
automating  the set up and it will require re-work later to use this workaround now.

We anticipate releasing updated errata this month, depending on how testing
goes. Thanks for the bug report!

fix confirmed.

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-2006-0278.html

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-2006-0284.html