fix(f3): try to make f3 participation stick to a single node (#12664)

Author: Stebalien

CHANGELOG.md

@@ -24,6 +24,7 @@
 - Fix issue where F3 wouldn't start participating again if Lotus restarted without restarting the Miner ([filecoin-project/lotus#12640](https://github.com/filecoin-project/lotus/pull/12640)).
 - Change the F3 HeadLookback parameter to 4 ([filecoin-project/lotus#12648](https://github.com/filecoin-project/lotus/pull/12648)).
 - Upgrade go-f3 to 0.7.1 to resolve Tipset not found errors when trying to establish instance start time ([filecoin-project/lotus#12651](https://github.com/filecoin-project/lotus/pull/12651)).
+- Try harder in the F3 participation loop to participate using the same lotus node ([filecoin-project/lotus#12664](https://github.com/filecoin-project/lotus/pull/12664)).
 - The mining loop will now correctly "stick" to the same upstream lotus node for all operations pertaining to mining a single block ([filecoin-project/lotus#12665](https://github.com/filecoin-project/lotus/pull/12665)).
 
 ## Deps

chain/lf3/participation.go

@@ -14,6 +14,7 @@ import (
 	"github.com/filecoin-project/go-f3/manifest"
 
 	"github.com/filecoin-project/lotus/api"
+	cliutil "github.com/filecoin-project/lotus/cli/util"
 	"github.com/filecoin-project/lotus/node/modules/dtypes"
 )
 
@@ -89,6 +90,11 @@ func (p *Participant) run(ctx context.Context) (_err error) {
 		}
 	}()
 
+	// Try to make send all requests to the same node. If a request fails, we'll switch nodes.
+	// This interacts with the FullNodeProxy, which is how we support multi-node setups by
+	// default.
+	ctx = cliutil.OnSingleNode(ctx)
+
 	var ticket api.F3ParticipationTicket
 	for ctx.Err() == nil {
 		var err error