juanma9613

Hi, In my conda environment I have cuda 11.1, so I was trying to follow @tae898 suggestion. But I get the following error, saying that my gcc version is incompatible,...

Hi @ckadner , @rafvasq, @Jooho , I wanted to ask you whether there's any task going on to improve the docs on HPA for modelmesh-serving? It's just not clear from...

I found the reason why this happens. It doesnt happen because the save and load but because the non deterministic implementation of the knn: knn voting system is not deterministic...

@N-Kingsley , I'm not a mantainer, but the way to address this at least for the /load endpoint is by adding the model versions you require in your config.pbtxt version...

@moroclash , how did you end up solving this issue? It's happening to me mainly because spot instances terminating and new instances being created in the middle of a model...

Hi, I'm indeed using go client v6 and reading code and docs from v6, maybe I sent the wrong one by error, could yo please tell me which link points...

@khaf, we observed the server side and indeed found some timeouts when the issue started. So, according to [this](https://docs.aerospike.com/reference/metrics#client_tsvc_timeout) it can be hotkey + too aggressive timeout in our client...

@khaf , I'm gonna do the following then to make sure my timeout of 3ms is respected in my app ``` enclosing_fn_with_3ms_timeout(): go call aerospike asyncronously (basepolicy.totaltimeout 100ms -> basepolicy.sockettimeout...

@khaf I was able to solve the issue by first limiting the openingConnections to 64, by also increasing the timeouts and finally and much more important: I had to stop...

@khaf , either what you said or there's an infinite error checking there. Unfortunately I found the fix of just not using the errors from the library for these error...