KeepAliveInterval sets an unknown gRPC option, so keepalive pings are never sent #38

What happens

BaseKurrentDBClient.__init__ maps KeepAliveInterval to the channel option grpc.keepalive_ms. gRPC does not know this option and ignores it. The gRPC option is grpc.keepalive_time_ms. Because of this, KeepAliveInterval and KeepAliveTimeout in the connection string have no effect, and the client sends no keepalive pings.

Steps to reproduce

1. Connect with ?KeepAliveInterval=5000&KeepAliveTimeout=5000.
2. Freeze KurrentDB with docker pause and start a read.
3. Run with GRPC_TRACE=http_keepalive GRPC_VERBOSITY=DEBUG.

Result: no keepalive pings are sent, and the read waits until KurrentDB responds again.

A second detail

With grpc.keepalive_time_ms set, the pings are sent. But with grpcio 1.84, an unanswered ping fails the connection only after grpc.http2.ping_timeout_ms, which defaults to 60 s. grpc.keepalive_timeout_ms did not shorten this in my test.

With grpc.keepalive_time_ms=5000, grpc.http2.ping_timeout_ms=5000 and grpc.http2.max_pings_without_data=0, 40 reads against a frozen KurrentDB failed at 9.7 s with ServiceUnavailableError: Stream removed (ping timeout). The client read normally again after the unpause. Normal use with 5 s pings for about 95 s, with idle gaps up to 45 s, caused no too_many_pings errors.

Proposed change

• Map KeepAliveInterval to grpc.keepalive_time_ms.
• Also set grpc.http2.ping_timeout_ms from KeepAliveTimeout, so that an unanswered ping fails the connection in time.
• Set grpc.http2.max_pings_without_data=0, so that pings continue while a call waits without data.

I'm opening a PR with this change.

Versions: kurrentdbclient 1.3.3, grpcio 1.84.0, KurrentDB 26.1.2, Python 3.13.