F5 LTM memory usage id 18.1% and busiest CPU remains between 6 to 13 %. But yesterday and today again I noticed a spike in busiest CPU and busiest cpu rises to 100% while average cpu was 17 at that time. Can you tell me what could be the problem? I have 2 vs of main traffic. 1 has a lot of traffic and CMP is enable on that one but another one has normal traffic and CMP is disable on that one because Server administrator was demanding exact round robin and not 2 requests on one webserver. Could that VS be producing this kind of spike? How can I trouble shoot it?

i mot sure that load on vs an cause such issue, mostly i prefer to check the current software release and HF, also seek the release note might you have some setting is not compatible with the current release

its 11.4.1 HF5. Yes load doesn't increase when i observe the spike and spike most of time reaches bw 90 to 100% although average is 17 at that time but which cpu goes this high and how to find this one.

It can be anything. Check the logs and look for repeating logs. There can be problems like nodes flapping, OS bug anything. show sys log ltm If you see something fishy open a support case.

how to check last logs so that I can check right at that time, command like tail -f. There is command show sys log ltm lines xx.

is it happening while you do something on the f5? are you creating UCS backups when the CPU spikes?

Busiest CPU jumps to 100% usage

17 Replies

Mahmoud_Eldeeb_
Cirrostratus
Dec 17, 2014
i mot sure that load on vs an cause such issue, mostly i prefer to check the current software release and HF, also seek the release note might you have some setting is not compatible with the current release
- Muhammad_Irfan1
  Cirrus
  Dec 17, 2014
  its 11.4.1 HF5. Yes load doesn't increase when i observe the spike and spike most of time reaches bw 90 to 100% although average is 17 at that time but which cpu goes this high and how to find this one.
kash_49328
Nimbostratus
Dec 17, 2014
It can be anything. Check the logs and look for repeating logs. There can be problems like nodes flapping, OS bug anything.

show sys log ltm

If you see something fishy open a support case.
- Muhammad_Irfan1
  Cirrus
  Dec 17, 2014
  how to check last logs so that I can check right at that time, command like tail -f. There is command show sys log ltm lines xx.
- kash_49328
  Nimbostratus
  Dec 18, 2014
  tail -f does work in F5 but it has to be in the bash not in tmsh so you have enter something like below. run util bash tail -f /var/log/ltm
shaggy
Nimbostratus
Dec 18, 2014
is it happening while you do something on the f5? are you creating UCS backups when the CPU spikes?
- Muhammad_Irfan1
  Cirrus
  Dec 19, 2014
  No I am not taking UCS backup which busiest cpu spikes. And whenever it spikes it spikes to 100% otherwise it doesn't spike. It does not happen very often. Average at the time of spike is 18%. Could this be the CMP disable at one VS out of my two VS.
- Muhammad_Irfan1
  Cirrus
  Dec 19, 2014
  During creating QKVIEW file I experienced the busiest cpu to 90 and average 20%. Why cpu is not equally divided? But get these spikes during times when I am not making backup.
- shaggy
  Nimbostratus
  Dec 19, 2014
  SOL15710: Large BIG-IP files may cause the csyncd service to spike CPU utilization - https://support.f5.com/kb/en-us/solutions/public/15000/700/sol15710.html disabled CMP could cause a single CPU to spike if something happened to that virtual server - for example, if an irule is applied and one request caused something to go awry, or if that virtual server suddenly had a high number of connections. You should open a support case.
shaggy_121467
Cumulonimbus
Dec 18, 2014
is it happening while you do something on the f5? are you creating UCS backups when the CPU spikes?
- Muhammad_Irfan1
  Cirrus
  Dec 19, 2014
  No I am not taking UCS backup which busiest cpu spikes. And whenever it spikes it spikes to 100% otherwise it doesn't spike. It does not happen very often. Average at the time of spike is 18%. Could this be the CMP disable at one VS out of my two VS.
- Muhammad_Irfan1
  Cirrus
  Dec 19, 2014
  During creating QKVIEW file I experienced the busiest cpu to 90 and average 20%. Why cpu is not equally divided? But get these spikes during times when I am not making backup.
- shaggy_121467
  Cumulonimbus
  Dec 19, 2014
  SOL15710: Large BIG-IP files may cause the csyncd service to spike CPU utilization - https://support.f5.com/kb/en-us/solutions/public/15000/700/sol15710.html disabled CMP could cause a single CPU to spike if something happened to that virtual server - for example, if an irule is applied and one request caused something to go awry, or if that virtual server suddenly had a high number of connections. You should open a support case.
Mahmoud_Eldeeb_
Cirrostratus
Dec 18, 2014
below link might be helpful for logs https://support.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/tmos-concepts-11-1-0/tmos_logging.html
Mahmoud_Eldeeb_
Cirrostratus
Dec 18, 2014
below link might be helpful for logs https://support.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/tmos-concepts-11-1-0/tmos_logging.html