daviesrob

Pull request #1049 adds more error reporting to BGZF. Could you give it a try, please? Hopefully it will provide more clues as to where the problem is.

That means it didn't find the expected GZIP signature at offset 1626705469. Was that at the start of the last request, or some distance into it? It should be possible...

The hexdump looks OK. There are, as far as I'm aware, no changes that would have made the 503 error appear. It could be a different manifestation of the existing...

Sorry, I've been diverted for the past few days by virtual conferences and an (unrelated) diversion into amplicon sequencing. Meanwhile, I've tried adding BGZF caching to tabix, and disappoiningly it...

Right, I think I may have a solution now. Could you give pull request #1053 a try, please? In my tests, it considerably reduced the number of http requests. Hopefully...

The first error was due to a bug - I've pushed an update to the pull request that should fix it. The second one looks more like a manifestation of...

Yes, I think we're reaching the limits of what we can do, apart maybe from adding even more logging. I guess if the failure rate is low enough now, you...

I doubt that it's a problem with zlib as that should be completely deterministic. It must either be GCS not always sending the same data back, or something randomly altering...

This could print lots of lines if the problem happens at the start of a long header. There's also no way of limiting the output if the header doesn't include...

That sounds rather slow. It may be a scaling problem with the length of your HiFi reads. We'll try running it on some and see if it's possible to reproduce...