Tech Tip: BIG-IP as Upgrade Facilitator

If you have machines behind your BIG-IP that are not load balanced, you still get a ton of benefits from their location. Virtualization is a big one, as is the number of metrics that are available in the BIG-IP.

This Tech Tip focuses on a simple way to get more out of your BIG-IP by simply putting servers behind it. Yep, we said that. Just put the servers behind the BIG-IP and get more.

The idea is simple. Upgrades cost downtime - generally a not insignificant amount of downtime. But what if you could reduce that downtime to zero? What if we said the chances were minuscule that even one customer would be effected? Well you can, if you have extra hardware.

 Assuming you’ve got your server to be upgraded behind a BIG-IP, here are some simple steps to upgrade:

•  If you’ve got another box that you can use for your upgrade, then place it behind your BIG-IP.
• Give the new box any old open IP address and hostname.
• Install everything the machine needs.
• Place the machine in a pool of its own – we’ll call it testPool.
• Create a new Virtual Node (we’ll call this testVIP) and assign testPool to be the default pool.
• Copy all of the relevant data from your current (production) server.
• Give your testers rights to testVIP, and run acceptance testing.
• Copy all of the relevant data again to get the changes in production since testing started.
• Add your new server to the current production pool.
• Change the state on the old server to “only active connections allowed” this will stop accepting connections and break persistent connections. In short, all new traffic will be routed to the new node. Because we did not allow persistent connections to stay with the old server, when they immediately try to reconnect the BIG-IP  will direct them to the new server.
• After all connections have ended with the old server, remove it from the current production pool.
• Decommission your old server in whatever manner is normal for your company.

The benefit here is the virtualization of BIG-IP. The Virtual Node address of your production server never changed, so users saw no issues. You did not have to take down the existing server, change a bunch of IP settings on the new server, then bring it online. You just add the new server to the pool and change the state of the old server.

Regular end users will not experience any downtime in this process, and persistent connections will reset and attach to the new server. You look like a hero for such a smooth upgrade, your boss will get congratulations on “pulling it off”. Everyone is happy, and the extra work involved is simply a setting change in your BIG-IP.