unpredictable API response for streaming keys/buckets

[cipy]

When performing a bucket or key streaming operation, the riak's API would in some cases:

• respond with the full set
• respond with a partial set, then {timeout}
• respond with an empty set

[binarytemple]

This is using the HTTP API

[cipy]

Can be seen in clusters where the data on disk greatly (i.e. at least 10 times) exceeds the available RAM dedicated to Riak.

[gglanzani]

I'm consistently getting partial set, then {timeout}. The workaround (I'm using the Python client) for our use case (delete keys), is to wrap the stream in a try/except block wrapped in a while block, and let the client reconnect.

[russelldb]

Are you setting the timeout high enough on the request?