NimbostratusHello everyone, can someone guide me in how to cutover an F5 system to new hardware? Assuming all configs have been loaded on the new system and HA, vlans and self-ip's have been created, what is the actual procedure for pushing the "magic button" to bring the new system online?
Thanks!
Toneman
MVPperhaps i don't understand the question but i don't believe there is one. you will need to disconnect the old device and connect the new one.
if you built a second box next to the old one and have a cluster then you can failover, is that what you are looking for?
NimbostratusI'm cutting over one HA pair to another. In the past I've removed the interfaces from the outgoing units and immediately assigned the interfaces to the new pair. Just wondering if this was the standard way of doing it.
MVPyeah, that sounds like the general idea. i would probably shutdown switch ports and un shut others. but it remains the general idea. perhaps perform a failover for some gratuitous ARPs.
CumulonimbusThis depends on a lot of other factors so will share a few ways I have cutover to new HA pairs in the past:
Cloned confit then re-IP’ed all Self-IP’s and disables all VS’s and VS Addresses. Bring system online and allow all monitors to bring up pools, do a check etc. To migrate disable VS Address on the original and enable on the new, likely to need an ARP and MAC table clear on the connected switches.
If using F5 GTM/DNS as well then do the same as above but re-IP all the VS’s as well then update GTM/DNS pools etc. This is the cleanest way to migrate because as long as the configuration is all working can manage migration per FQDN with each application tested etc.
Finally the direct cut over as you have stated above. Here I have often re-IP’ed the Self-IPs on the server side VLANs to allow the monitors to work and pools to come up before migration takes place and speeds up the cutover. Again clear ARP and MAC address tables on the connected switches during cutover. Have seen this method used so well had very little disruption but that bepend on the applications and your individual setup.
Hope this helps a bit.
It’s also possible to extend the current cluster from for example 2 to 4 members. Make sure all members are in sync. Than make one of the new members the active one. After that remove the old members from the cluster. I’ve done this before replacing two 2200s for 4000s appliances.
