Previously when the dupserver gearmand instance fails the retry would override current_client with current_client_dup leading to results being missed by the non-dupservers.
We've changed it so that create_client no longer sets current_client and the logic for setting this has been pushed to the consumer of the create_client function. This means we can delete create_client_dup; this also means that the timeout is now set for dupserver clients.
This may also fix #111 as it looks similar in behaviour
Previously when the dupserver gearmand instance fails the retry would override
current_client
withcurrent_client_dup
leading to results being missed by the non-dupservers.We've changed it so that
create_client
no longer setscurrent_client
and the logic for setting this has been pushed to the consumer of thecreate_client
function. This means we can deletecreate_client_dup
; this also means that the timeout is now set for dupserver clients.This may also fix #111 as it looks similar in behaviour