Closed sylwiaszunejko closed 2 months ago
If we have TCP keepalive, why do we need the application level OPTIONS as well? (and vice versa of course).
Should the fix be sent to https://github.com/gocql/gocql first?
Should the fix be sent to https://github.com/gocql/gocql first?
Not sure about that. It looks like the upstream is not actively maintained, and this change is not something that might make merging upstream difficult in the future.
Should the fix be sent to https://github.com/gocql/gocql first?
Not sure about that. It looks like the upstream is not actively maintained, and this change is not something that might make merging upstream difficult in the future.
I agree. Normally, such general changes are better to go first to the upstream and then get them during the upstream merge, but in the current situation of upstream it can wait there unnoticed for months.
@sylwiaszunejko - what's the latest here?
@avelanarius could you review it?
Now we are sending OPTIONS message on every connection every 5 seconds.
This PR changes that to 30 seconds (this is the number used in python-driver - https://github.com/scylladb/python-driver/blob/7cdc1ca25ffa2194a0772e46ccd89645f8514c1b/cassandra/cluster.py#L893) and enables the keepalive - with 15s by default (matches Golang's default).
Fixes: #162