Closes: #8575
This PR aims to fix the `LastCommitRound can only be negative for initial height 0` issue we see in the e2e tests by initializing the `state` object before starting the receive routines in the consensus reactor. This is somewhat inelegant, but it should fix the issue.
* Fix lock sequencing in socket client request tracking.
It is not safe to check base service state (IsRunning) while holding the lock
for the client state. If we do, then during shutdown we may deadlock with the
invocation of the OnStop handler, which the base service executes while holding
the service lock.
* Enqueue pending requests before sending them to the server.
If we don't do this, the server can reply before the request lands in the
queue. That will cause the receiver to terminate early for an unsolicited
response. So enqueue first: This is safe because we're doing it in the same
routine as services the channel, so we won't take another message till we are
safely past that point.
* Document what we did.
* Fix socket paths in tests.
In #3435 we allowed this timeout to override the global write timeout.
But after #8570 this meant we were applying a shorter timeout by default.
Don't do the patch if the timeout is already unlimited.
This is a temporary workaround; in light of #8561 I plan to get rid of this
option entirely during the v0.37 cycle, but meanwhile we should keep existing
use more or less coherent.
* rpc: rework timeouts to be per-method instead of global
Prior to this change, we set a 10-second global timeout for all RPC methods
using the net/http Server type's WriteTimeout. This meant that any request
whose handler did not return within that period would simply drop the
connection to the client.
This timeout is too short for a default, as evidenced by issues like [1] and
[2]. In addition, the mode of failure on the client side is confusing; it
shows up as a dropped connection (EOF) rather than a meaningful error from the
service. More importantly, various methods have diffent constraints: Some
should be able to return quickly, others may need to adjust based on the
application workload.
This is a first step toward supporting configurable timeouts. This change:
- Removes the server-wide default global timeout, and instead:
- Wires up a default context timeout for all RPC handlers.
- Increases the default timeout from 10s to 60s.
- Adds a hook to override this per-method as needed.
This does NOT expose the timeouts in the configuration file (yet).
[1] https://github.com/osmosis-labs/osmosis/issues/1391
[2] https://github.com/tendermint/tendermint/issues/8465
