Forum Discussion
Simon_83666
Nimbostratus
Nimbostratussyslog-ng & alertd CPU usage
We're having an issue with our new LTM 6400 box. With the following iRule enabled for logging, the syslog-ng & alertd processes are taking up almost 100% of one CPU during a performance testing...
Nicolas_Menant
Employee
Employeelogging and clock are really heavy for cpu. Do you have an idea about how much request/sec you have going through your BIGIP ?
If you want to have an estimation you can use the timing command [url]http://devcentral.f5.com/wiki/default.aspx/iRules/timing.html[/url], [url]http://devcentral.f5.com/Default.aspx?tabid=63&articleType=ArticleView&articleId=123[/url]
I did some testing on a 6400 Platform with v9.4.3 and here is what i found when i did my tests:
[code]
iRule:
when HTTP_REQUEST timing on {
}
Result:
+-> HTTP_REQUEST 25190 total 0 fail 0 abort
| Cycles (min, avg, max) = (664, 683, 9969)
when HTTP_REQUEST timing on {
set actual_time [clock seconds]
}
Result:
+-> HTTP_REQUEST 69031 total 69031 fail 0 abort
| Cycles (min, avg, max) = (12461, 12601, 94735)
Command name average cpu cycles CPU usage / request max requests / second to take (100% cpu with the iRule ONLY!)
log 10064 0,000504871% 198070
clock 12601 0,000632142% 158192
[/code]
HTH
