Closed igorwwwwwwwwwwwwwwwwwwww closed 7 years ago
Just rolled a new release (1.1.0) with this in it. Thanks again.
It looks like this patch may have introduced a memory leak. From some testing, it looks like faraday makes use of timeout
, which creates a thread per call. Some rudimentary memory profiling points to this as the source of un-freed memory. I am digging into this and will see if I can find out more about the strange error we saw with the http
gem.
@igorwwwwwwwwwwwwwwwwwwww oof, thanks for the report. Let us know how it goes, and we'll take a look from our end to see if we can get some better behavior out of faraday.
for some set of large responses, the http gem was raising an error:
i am not completely sure what the conditions for that failure are, but request body size seems to be correlated with this occurring consistently.
switching to faraday fixed the problem.
another thing worth noting is that this error was not being rescued anywhere, which made it quite difficult to spot. that is something that should be addressed regardless of this patch.