Set up Amazon EC2 instance as backup for collecting relay descriptors
metrics-db on yatei downloads the current consensus and all referenced descriptors in an hourly cronjob. Until a few weeks ago we had weasel's host collect relay descriptors, too, which we imported into yatei's metrics-db. But that host is gone. If yatei becomes unavailable for an hour or even longer, we'll never download those descriptors. That's bad, because we rely on relay descriptors to be complete for ExoneraTor, metrics tarballs, statistical analyses, etc.
We should run a second metrics-db instance in a different place than where we host yatei, configure it to download relay descriptors once per hour, rsync its output to yatei, and import those files into the metrics-db instance on yatei.
Andrew suggested an Amazon EC2 instance. That host:
can be a Small Instance with 160 GB of local instance storage,
needs to run 24/7, and
needs a sysadmin who's not me (of course, I'm going to operate the metrics-db service).