Here is the straight dope: We were getting everything ready for a move to Hurricane Electric's FMT2 facility, which is like two miles down the road from FMT1. This was all to be ready about a week after the upgrade announcement. FMT2 is a modern build-out, which we consider to be a great improvement over FMT1's aging power and space restrictions. The strategy involved us getting a 10G wave on fiber they have between buildings, which places our two distinct networks on the same layer2 network. Meaning no IP swaps, no disruption in IP failover support, private networking working as expected, etc. Essentially this has created one large LAN despite hardware and Linodes existing in both facilities. Over the course of 3 months or so, we would perform a few phases of migration queues + physically moving and upgrading hardware from FMT1 into FMT2, and then continuing with the next phase, rinse, repeat. Doing it in phases was largely a host-resource game.
This was all staged and ready to go, but we discovered this link wasn't as redundant as we had believed. We asked questions, and got answers from HE. There are scenarios where the wave between the two facilities could break, and that news put a stop to everything while we investigated how to implement absolute redundancy between the two facilities in the event the link failed. We have a solution, but want to greatly shorten the time we're relying on this bridged network, which means altering the FMT1 to FMT2 transition timeline. This is where we stand now.
Things could change, but we're now thinking this will be: a few weeks heads-up during which the upgrade-button would simultaneously upgrade you AND migrate you to FMT2. Those who don't perform the upgrade by the deadline(s) would have their host shut down, the host de-racked and physically moved to FMT2 and then powered back on. Of course we would have everything prepared in FMT2 to receive these machines and essentially it would be: a shutdown, unplug host, transport it to FMT2, rack it and plug it into already-ran cables, and then power it on. Logistics so far indicate the downtime for that scenario is minimal, 60-120 minutes, performed in the middle of the west-coast night. Over the course of a few days we'd be out of FMT1 completely.
We've already deployed an impressive amount of capacity in FMT2, however this shortened timeline and elimination of phases means upgrade+migrate-to-fmt2 availability will be on a first-come-first-served basis. We've also been working on redirecting host builds from other facilities to fmt2 to augment fmt2's host resources. However, we can't guarantee everyone that wants to move/upgrade early will be able to, under this scenario.
There are still a few loose ends, and of course this could all change, but that's where we are. The logistics of this type of endeavor are quite complicated as you can imagine, and we want to do this with the least amount of stress, downtime, and risk as possible. We hope to nail down the timeline and enable the upgrade button very soon.