Open zimnx opened 1 year ago
master
When connections to cluster are proxied, and all of them lands on the same proxy replica it's impossible to do a rolling restart of a proxy, even when there's always a second proxy replica able to handle the traffic.
Instead throwing an exception and failing entire request, driver should attempt to establish connection once (or with configurable amount of retries) and fail only if it's not possible. https://github.com/scylladb/java-driver/blob/scylla-4.x/core/src/main/java/com/datastax/oss/driver/internal/core/cql/CqlRequestHandler.java#L311-L321
Requests are failing for noticeable amount of time until pool is filled with at least one connection.
Leaving this one only for SNI Proxy case that should support rolling restart.
I'm not sure SNI proxy is an interesting use case.
Right, hence we set it to backlog and lowest priority for now.
What version of driver are you using?
master
What did you do?
When connections to cluster are proxied, and all of them lands on the same proxy replica it's impossible to do a rolling restart of a proxy, even when there's always a second proxy replica able to handle the traffic.
What did you expect to see?
Instead throwing an exception and failing entire request, driver should attempt to establish connection once (or with configurable amount of retries) and fail only if it's not possible. https://github.com/scylladb/java-driver/blob/scylla-4.x/core/src/main/java/com/datastax/oss/driver/internal/core/cql/CqlRequestHandler.java#L311-L321
What did you see instead?
Requests are failing for noticeable amount of time until pool is filled with at least one connection.