Forum Discussion

Piotr_Lewandows's avatar
Piotr_Lewandows
Icon for Altostratus rankAltostratus
May 17, 2018

How to rebuild vCMP guest VM - Viprion deployment

Hi,

I wonder if anyone had issue vCMP guest VM failure on one slot. I have three slot Viprion with guests set to all slots. After upgrade guest started but the guest VM on slot 2 failed. So now guest is running only on two slots. Slot 2 VM seems to be completely dead:

  • No way to SSH using this slot member IP
  • ssh slot 2 from slot 1 not working
  • vconsole [guest name] 2 from host not working

I was not able to test 

tmsh reboot slot 2
from guest slot 1 (Primary) but I highly doubt it will work. Sure I will try reboot guest chassis (like 
tmsh reboot slot all
) but what if slot 2 VM will not boot again?

Any proven procedure?

My idea is to:

  • Change guest state to configured
  • Disable guest on slot 2

  • Change state to provisioned - according to info collected:

    • VM will be deleted
    • vdisk will be preserved

  • I think vdisk on slot 2 should be removed - hope can be done via host GUI vCMP ›› Virtual Disk List

After clean up:

  • Change state to configured
  • Enable guest on all slots (original setting)

  • Change state to provisioned, I assume that:

    • VM will be created
    • vdisk will be copied from primary slot

  • Change state to deployed

Should it work? Any better/safer way?

Piotr

devops
guest vm
LTM
rebuilding
slot
vcmp
viprion
  • dragonflymr's avatar
    dragonflymr
    Icon for Cirrostratus rankCirrostratus
    May 21, 2018

    Hi,

     

    I would like to share official F5 procedure for rebuilding guest VM on given slot.

     

    1. Disable guest (change state to Configured)
    2. Remove virtual disk on affected slot.
    3. Change guest configuration on host so that guest is not started on slot which got virtual disk removed (this is to ensure that this blade will not become primary and will not perform blank config synchronization to other blades).
    4. Affected slot should removed from the list of "Allowed on slot list" option for the guest.
    5. Once guest is fully started on all slots with exception to slot which got virtual guest removed - adjust guest configuration so that missing slot is added again to guest's "allowed on slot list". It can be done without shutting down guest.

    Point 3 above is very important. After removing vdisk for given slot and setting guest to provisioned new empty vdisk is created and then software installation as specified by Initial image is performed - no configuration is copied at this stage.

     

    During deployment phase system replicates the configuration from primary slot and as well performs software installation to match version active on primary slot.

     

    If for some reason slot with new vdisk containing just Initial image installed and no configuration is selected as primary (rare but possible) empty configuration will be pushed to other slots (being secondary at this time).

     

    As an result all original configuration will be lost for guest.

     

    Piotr

     

  • youssef1's avatar
    youssef1
    Icon for Cumulonimbus rankCumulonimbus
    May 17, 2018

    Hi Piotr,

     

    I worked a lot on the viprion part, but never encountered this kind of problem. the simplest test is indeed a reboot before engaging in complicated procedures. Disable guest on slot 2 and remove vguest associate then enable it again in slot 2 will built the guest on this slot... I think it could work.I would have done the same...

     

    as far as I'm concerned I only use one slot per guest clustered and it allows me to manage my high availability.

     

    Did you generate a qkview on the Host? in order to check if you have an unusual log? and keep it for support in case you want answers on this issue.

     

    your second alterative could also correct the problem. but we must make the simplest first make a reboot of the host.

     

    I advise you to open a ticket with the support. this behavior is not normal and from what we see it is clearly not a problem of configuration ...

     

    Regards