Currently the parsing team is running a bunch of big tests on Promethium. Promethium was a one-off bare-metal-in-the-cloud experiment; it's out of warranty and we don't want to move any of those bare-metal hacks into the new Neutron labs deployment, eqiad1.
Subbu tells me that ideally he'd like a one-for-one replacement (one big VM) but could possibly live with a cluster of smaller ones.
Prometheum's specs are:
32Gb Ram
400 Gb disk
12 cores
It seems silly to build that out in eqiad only to have to immediately copy it over to eqiad1, so let's wait and have this be an early use of eqiad1 once it's ready to go. Tagging @bd808 in case he thinks this has budget implications. As a single big VM it's a big ask but the actual resource implications aren't outrageous for an internal project request.