David Acevedo
David Acevedo
> Our backend code creates many sessions/resources They do [recommend](https://github.com/aio-libs/aiobotocore/issues/928#issuecomment-1118945320) keeping a single persistent session for long-running processes like servers. We did this a month ago and saw slightly better...
@zferentz tbh I'm glad you reported this as we were also seeing some performance shenanigans but weren't really sure whether event loop skews due to blocking calls played a part...
I'm not familiarized with the inner workings of boto, but I found this test case: https://github.com/boto/botocore/blob/develop/tests/unit/test_retryhandler.py#L267 Perhaps they intend to retry read timeouts but not connect timeouts, which is weird...
Not stale
Not stale
> Yeah, the export of the commonjs module is a Mixpanel library instance. It won't work to call individual methods without any binding to their instance. So would you suggest...
note: It works well using the tar.gz `npm i --save https://github.com/cloudinary-devs/gatsby-source-cloudinary/archive/19a2115fdf876c55ff74756e6dfd8d775bad636c.tar.gz` npm supports both, so it would be worth documenting this behavior in case this becomes a `won't fix`
Any updates? This would be useful to report opentelemetry metrics
Same problem 6 years later :disappointed: I'm using amazon-kinesis-client 1.13.3 with dynamodb-streams-kinesis-adapter 1.5.3 This is especially annoying in combination with the already spammy MultilangDaemon