Forum Discussion

Jason_L_40779's avatar
Jason_L_40779
Icon for Nimbostratus rankNimbostratus
May 01, 2011

LTM CPU utilization Question

I have a question and maybe someone can help me understand CPU utilization. I have a pair 0f 6900's (HA Pair) running 10.2.0 1755, which are provisioned as nominal Because of this, the CPU is generally right around 10%. I believe the OS reserves 10% just for management, hence the reason why it's running around 10%. Every Sunday around 4AM, I see my CPU utilization on both my active and standby unit do the same thing. They spike around 20-40% utilization for about 15-20 minutes then go back to normal. At that same time, I do not see any increase in thoroughput ssl transactions or anything according to dashboard or any of the statistics.

 

 

It just appears to be just the CPU utilization and it only happens on sunday's at 4AM. . Being both units are doing the same thing at the same time, I don't think it's an issue with one box in particular.

 

 

Is there any jobs, crons or anything that run on these appliances out of the box that could cause the CPU to do this on a weekly basis? Or if not, I'm wondering if this could be related to patches or some multicast going on the network at that time every week. I'm not seeing anymore traffic through the unit, but could it be alot of chatter on the wire it's listening to?

 

 

A couple things to note. I do not have a default route due to all one armed load balancing. I keep it simple and let the network do the routing.

 

 

Also, I have a seperate fiber cable plugged into each unit directly (not through a switch) for synch traffic and mirroring persistence. I only allow the synch vlan over that link.

 

 

Both unit's CPU graphs are identical each week. I'm just curious what could cause this..

 

 

I would appreciate any feedback anyone may have or experience with this...

 

 

Thanks,

 

 

Jayson

 

  • Hamish's avatar
    Hamish
    Icon for Cirrocumulus rankCirrocumulus

     

    @4AM it's more than likely going to be the log rollover... Which means an amount of gzip'ing going on... Which will of course spike the CPU. More logs means more compressing and more CPU consumed.

     

     

    Checkout SOL7115 (http://support.f5.com/kb/en-us/solutions/public/7000/100/sol7115.html) for an overview of the log rollover and affecting the timing (e.g. changing from daily, to weekly to monthly).

     

     

    H
  • thanks hamish. however, This event happens only weekly on both boxes identically and the log files rotate daily. I did find the entry in /usr/share/defaults/config/templates/logrotatecommon.tmpl. set to daily. So my logs rotate daily. Is there something else that happens with the logs weekly that i'm not aware of? I still consider myself newer with LTM so please bear with me..

     

     

    This was the entry in the logrotatecommon.tmpl.

     

    rotate log files daily

     

    daily

     

     

  • According to the SOL, this is the file where you would make the changes...Unless i'm missing something else. When you say config file... is there somewhere else or would this would be the file?

     

     

    Using a text editor, open the /usr/share/defaults/config/templates/logrotatecommon.tmpl file

     

     

     

    This is my file...

     

     

    Use the bigpipe shell utility to make changes to the system configuration.

     

    @perl_begin

     

    my $helpLine = "UNIX_CONFIG_LOGROTATECOMMON_REPLACE_TEMPLATEHELP";

     

    print "\ For more information, see $helpLine.\n";

     

    @perl_end

     

    rotate log files daily

     

    daily

     

    logrotate_enhance controls the log files life cycle

     

    rotate UNIX_CONFIG_LOGROTATECOMMON_DEFINE_BACKLOGS

     

    create new (empty) log files after rotating old ones

     

    create

     

    compress

     

    It will override the --force and daily, weekly... options

     

    to rotate logs only when log file reaches certain size.

     

    UNIX_CONFIG_LOGROTATECOMMON_REPLACE_INCLUDE

     

  • It looks like it's 2 crons that are running on the boxes at 4AM each sunday. One is doing a RAID check, the other is doing a SSL cert check. That would explain it. . Thanks for the input.