Forum Discussion

Krypto50_190970's avatar
Krypto50_190970
Icon for Nimbostratus rankNimbostratus
Mar 27, 2015

No memory allocated to TMM processes other than TMM0.0

I'm running 11.6.3 HF-Engineering hotfix and I'm wondering if I'm running into a bug. The output below is for a 4000 with LTM and ASM provisioned.

 

I only see memory allocated to TMM0.0. I see no memory allocated to TMM0.2, TMM0.4 or TMM0.6. Shouldn't I see memory allocated to all of the TMM processes?

 

show /sys tmm-info

 

Sys::TMM: 0.0

Global TMM Process Id 16639 Running TMM Id 0 TMM Count 1 CPU Id 0

 

Memory (bytes)

 

Available 6.1G Used 521.4M

 

CPU Usage Ratio (%) Last 5 Seconds 2.55 Last 1 Minute 2.46 Last 5 Minutes 2.76

 

Sys::TMM: 0.2

Global TMM Process Id 16639 Running TMM Id 1 TMM Count 1 CPU Id 2

 

Memory (bytes)

 

Available 0 Used 0

 

CPU Usage Ratio (%) Last 5 Seconds 2.31 Last 1 Minute 2.31 Last 5 Minutes 2.63

 

Sys::TMM: 0.4

Global TMM Process Id 16639 Running TMM Id 2 TMM Count 1 CPU Id 4

 

Memory (bytes)

 

Available 0 Used 0

 

CPU Usage Ratio (%) Last 5 Seconds 2.64 Last 1 Minute 2.48 Last 5 Minutes 2.78

 

Sys::TMM: 0.6

Global TMM Process Id 16639 Running TMM Id 3 TMM Count 1 CPU Id 6

 

Memory (bytes)

 

Available 0 Used 0

 

CPU Usage Ratio (%) Last 5 Seconds 2.54 Last 1 Minute 2.93 Last 5 Minutes 2.82

 

4 Replies

  • I see the same on my Viprion running 11.6 HF4. I would say it's a bad behaviour but apparently (still) not seen by F5 as a bug (cannot find anything related on AskF5)
  • Either a bug, or perhaps someone intentionally configured the system to use a single TMM processor? Some severe system issues can be resolved by configuring the system to use a single TMM instance...

    Could you please paste the output for command

    tmsh list sys db provision.tmmcount
    , as well as for
    tmsh list sys db provision.tmmcountactual
    ?

  • The answer I received from F5 support is that this is expected behavior. With CMP, TMM is now a single process with multiple threads. All memory is allocated to the TMM process which manages the memory for the different threads. I'm still unclear on why the memory shows up under TMM0.0. I don't have access to the system right now but here is what I can get from the qkview file.

     

    list /sys db all-properties

     

    sys db provision.tmmcount { default-value "0" scf-config "false" value "0" value-range "integer min:0 max:1024" } sys db provision.tmmcountactual { default-value "0" scf-config "false" value "4" value-range "integer min:0 max:1024" }

     

    • Hannes_Rapp's avatar
      Hannes_Rapp
      Icon for Nimbostratus rankNimbostratus
      In case of vCMP, it makes sense, but do you have any vCMP Guests configured? For example, I have an F5 appliance (5250v) that has 8 vCMP Guests. When issuing the 'tmsh show sys tmm-info' command on any of the Guests systems, or the vCMP Host itself, I only see one TMM instance, even on the guests that are configured to use up 2 cores instead of 1. In case you have vCMP configuration in place, seeing the provisions only on TMM0.x is as expected - nothing to worry about. If you don't have any vCMP configuration in place, you do have something to be concerned about. Perhaps this arcticle will help you: https://support.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/vcmp-administration-appliances-11-5-0/1.html. Cheers!