Replacing LTMs in GTM Server List

Forums Technical Forums
1

I am currently working towards replacing two pairs of HA LTMs.   Each pair is configured in GTM (separate appliances) as one Server with two Devices.   The GTM configuration can be represented as

DC1 ->ServerDC1->LTMDC1-1 + LTMDC1-2

DC2->ServerDC2->LMTDC2-1 + LTMDC2-2

To prevent losing virtual server and pool objects, I believe the best way forward is to replace the LTMs one-by-one in the GTM Server configuration:

1) shutdown LTMDC1-1

2) delete device LTMDC1-1 from server Server DC1

3) assign OLD LTMDC1-1 self-IPs to NEW LTMDC1-1

4) add NEW LTMDC1-1 as a device to ServerDC1

5) big3d_install & bigip_add NEW LTMDC1-1

6) Rinse and repeat: LTMDC1-2, LTMDC2-1, LTMDC2-2

Am I on the right track here?

--Jason

2

The method is correct, however you can’t proceed without loosing virtual servers because they exist within the server you want to delete. So in case you can’t use virtual servers auto discovery, you have to create them again. I suggest you export them before deleting the server using:

tmsh list gtm server YOURSERVER virtual-servers one-line

Keep the output in notepad. And after adding the new server (I assume with the same name), add the lines using:

tmsh load sys config from-terminal merge

Regarding pools, you will not loose them, however I’m pretty sure you have to remove from them any references to the virtual servers you need to remove

3

Thank you for the granular back up suggestion.  I’ve put that to use extensively while migrating the LTMs to their new platform.  I was planning on keeping mulitple backups within reach during this reconfiguration.

I do not intend to delete the servers -- ServerDC1, ServerDC2 -- however.  I intend to delete the devices under the Servers.  Here’s what the config looks like from 'tmsh list gtm server ServerDC1`

gtm server ServerDC1 {<br>    addresses {<br>        10.10.10.1 {<br>            device-name LTMDC1-1<br>        }   <br>        10.10.10.2 {<br>            device-name "LTMDC1-1 Peer"<br>        }<br>    }<br>    datacenter DC1<br>    devices {<br>        LTMDC1-1 {<br>            addresses {<br>                10.10.10.1 { }<br>            }<br>        }<br>        "LTMDC1-1 Peer" {<br>             addresses {<br>                 10.10.10.2 { }<br>             }<br>         }<br>    }<br>    monitor bigip<br>    product bigip<br>    virtual-servers {<br>    /Common/virtual1 {<br>        destination 1.1.1.1:https<br>        }<br>    /Common/virtual2 {<br>        destination 1.1.1.2:https<br>        }<br>    /Common/virtual3 {<br>        destination 1.1.1.3:https<br>        }<br>and so on....  

This suggests to me that I can delete one device, and the other device will continue to feed the same VSs to GTM.  Afterall, if the active member of a LTM HA pair crashes and burns, and the standby takes over, the standby will feed the same VS status data to GTM, continuing to populate the Server’s virtual-servers configuration.   Yes?  The thought that triggered posting this note was: Will GTM act differently between two different scenarios, 1) iquery comms are down from one of a pair of HA devices, and 2) a device, one of two in an HA pair, has been administratively deleted?   I suppose should the latter scenario cause the virtuals under ServerDC1 to be deleted, I will restore them from backup.

4

@SalishSeaSecurity Assuming these DCs are active/active meaning you always send requests to both locations based on a GTM load balancing method and 1 site can handle all traffic I would go the route of replacing 1 DC at a time. So you suspend DC1, replace both devices at that DC, perform some testing of the replacement devices, and finally do the same for DC2 after DC1 is back as active in the GTM. I see no reason to overcomplicate this by trying to removed 1 device at a time from each DC if either DC is capable of handling all traffic.