Megan K

I can get the cache logs for you in a bit. The other modifications we made were just to upload the syncd packages to Amazon AWS. Can't imagine that's terribly...

I can give a better report when I get back to my desk, but it's the call to fetch the package json that was failing. I believe the x-pypi-last-serial header...

Okay, here's the full report: About 12 days ago, our mirror cron job started dying on every run. The logs all showed this exact message as causing the failure: `StalePage:...

@di can anyone purge the cache? If so I can add that to our codebase so we stop busting caches unneededly.

I can make a PR for this as well.

@dstufft is client purging something we can do in the short term until PyPI is fixed? I'm just offering the perspective of a user here: we need bandersnatch to work...

I still think the cost should be quite small for a hashtable-style index. Yeah it is a large change unfortunately. I think storing instances directly and then making a new...

The way I solved that was by doing away with eager/lazy loading completely. Instead, each property was just a lookup each time - but thanks to the hashmap indexes, the...

>For example, let's say we to filter the posts. If we do eager loading we only have to filter entire posts once no matter how many users record we have,...

I'm short on time for the next day or two, but I wanted to briefly respond. First, yeah, indexes and filtering/storing instances are two distinct but related problems. We've now...