The grpc-proxy test logic was assuming that the context associated to client is closed,
while in practice all tests called client.Close() without explicit context close.
The current testing strategy is complicated 2 fold:
- grpc proxy works like man-in-the middle of each Connection issues
from integration tests and its lifetime is bound to the connection.
- both connections (client -> proxy, and proxy -> etcd-server) are
represented by the same ClientV3 object instance (with substituted
implementations of KV or watcher).
The fix splits context representing proxy from context representing proxy -> etcd-server connection,
thus allowing cancelation of the proxy context.
This change makes the etcd package compatible with the existing Go
ecosystem for module versioning.
Used this tool to update package imports:
https://github.com/KSubedi/gomove
Problem Observed
----------------
When there is no etcd process behind the proxy,
clients repeat resending lease grant requests without delay.
This behavior can cause abnormal resource consumption on CPU/RAM and
network.
Problem Detail
--------------
`LeaseGrant()` uses a bare protobuf client to forward requests.
However, it doesn't use `grpc.FailFast(false)`, which means the method returns
an `Unavailable` error immediately when no etcd process is available.
In clientv3, `Unavailable` errors are not considered the "Halt" error,
and library retries the request without delay.
Both clients and the proxy consume much CPU cycles to process retry requests.
Resolution
----------
Add `grpc.FailFast(false))` to `LeaseGrant()` of the `leaseProxy`.
This makes the proxy not to return immediately when no etcd process is
available. Clients will simply timeout requests instead.
Piotr Tabor