There’s plenty of info out there about node failure in couchbase, but I can’t find in any of it an explanation on the availability of view indexes when a node is down.
Let’s say I have a 3 node cluster. After 1 node goes down and before it is failed over, the documentation says that roughly 1/3 of the get/set requests will fail since the documents are spread more or less evenly across the cluster. What is the expected behavior of view queries in this situation? Are 1/3 of the leafs in the index missing? If so, does the value of the stale parameter have any effect on this behavior?
In my experience with my cluster, when a node fails there are 0 get/set operations, not 1/3 less than normal. My first instinct is that this is due to secondary index failures. I don’t intend on purposely taking one down to test this, nor do I have spare machines lying around to make a test cluster. So, I figured a good first step would be to ask.