Closed mattjala closed 3 months ago
It seems that h5pyd is trying to read more bytes than was actually sent. See this PR: https://github.com/HDFGroup/hsds/pull/331 - that should fix the issue.
It seems that h5pyd is trying to read more bytes than was actually sent. See this PR: HDFGroup/hsds#331 - that should fix the issue.
This fixed the slowdown - this PR is ready for review now.
This currently has a massive performance issue with field reads. Reading 144 bytes in a field selection takes over a minute. The stall occurs when
rsp.iter_content
is invoked inbase.py
. Other calls that try to directly interact with the response, likersp.text
orrsp.content
have a similar delay.I tried manually specifying the
utf-8
encoding for the response, since apparently having to try a bunch of encodings when one isn't specifying can sometimes cause slowdown like this, but it had no effect.