Planet Migration Shakeout

This note is intended for Planet and its audience, to let you know that while we’re mostly up and running, we’ve found a few feeds that aren’t getting pulled in consistently or at all. I’m not sure where the problem is right now – for example, Planet reports some feeds as returning 403 errors, but server logs from the machines those feeds live on don’t show those 403s as having ever been served up. A number of other feeds show Planet reporting “internal server errors”, but again, no such errors are visible elsewhere.

Which is a bit disconcerting, and I have my suspicions, but I won’t be able to properly dig into this stuff for a few days. Apologies for the degraded state of the service, and I’ll report back with more information as I find it. Tracking bug is #1338588.

Update: Looks like it’s a difference of opinion between an old version of Python and a new version of TLS. I expect this to be resolved Monday.

Second update: I do not expect this to be resolved today. The specific disagreement between Python and TLS describes itself as the less-than-helpful SSL23_GET_SERVER_HELLO:tlsv1 alert internal error whose root cause can be found here; HTTPlib2 does not support SNI, needed to connect to a number of virtually-hosted blogs here in modernity, and it will take some more extensive surgery than expected to get Planet back on its feet.

Third update: The solution we have for this problem is to excise some outdated but vendored-in dependencies on Planet and move it to a recent version of Python. The combination of those things resolves this in staging, but it will take a few days before we can move this to production.