From 4a22692b433888355366261b07c31fcf9c4f58aa Mon Sep 17 00:00:00 2001
From: Mark Shabanov <mshabanov@openintegration.inc>
Date: Thu, 28 May 2026 12:06:57 +0300
Subject: [PATCH 1/2] prevent infinity wait for ALL_LOGS_RECRUITED on remote DC
 TC when there are no tlog processes in it  but usableRegions=2

---
 fdbserver/datadistributor/DataDistribution.cpp | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

diff --git a/fdbserver/datadistributor/DataDistribution.cpp b/fdbserver/datadistributor/DataDistribution.cpp
index b251cae0d51..f6aa29b94e9 100644
--- a/fdbserver/datadistributor/DataDistribution.cpp
+++ b/fdbserver/datadistributor/DataDistribution.cpp
@@ -299,6 +299,14 @@ Future<Void> StorageWiggler::finishWiggle() {
 	    });
 }
 
+Future<Void> waitForAcceptingCommits(Reference<AsyncVar<ServerDBInfo> const> db) {
+	TraceEvent("DDWaitForAcceptingCommitsStart").log();
+	while (db->get().recoveryState < RecoveryState::ACCEPTING_COMMITS) {
+		TraceEvent("DDWaitForAcceptingCommits").detail("RecoveryState", (int)db->get().recoveryState);
+		co_await db->onChange();
+	}
+}
+
 Future<Void> remoteRecovered(Reference<AsyncVar<ServerDBInfo> const> db) {
 	TraceEvent("DDTrackerStarting").log();
 	while (db->get().recoveryState < RecoveryState::ALL_LOGS_RECRUITED) {
@@ -2900,7 +2908,12 @@ Future<Void> dataDistribution(Reference<DataDistributor> self,
 				                                self->configuration,
 				                                self->remoteDcIds,
 				                                Optional<std::vector<Optional<Key>>>(),
-				                                self->initialized.getFuture() && remoteRecovered(self->dbInfo),
+				                                // In multi-region configurations DD only needs the cluster to
+				                                // reach ACCEPTING_COMMITS before continuing startup. Waiting for
+				                                // ALL_LOGS_RECRUITED can hang during degraded failover scenarios
+				                                // such as when the primary DC is down.
+				                                self->initialized.getFuture() &&
+				                                    waitForAcceptingCommits(self->dbInfo),
 				                                zeroHealthyTeams[1],
 				                                IsPrimary::False,
 				                                processingUnhealthy,

From b363e116727ca924b0febb97be667661e778b9cf Mon Sep 17 00:00:00 2001
From: Mark Shabanov <mshabanov@openintegration.inc>
Date: Fri, 29 May 2026 10:44:37 +0300
Subject: [PATCH 2/2] applied the clang format to DataDistribution.cpp

---
 fdbserver/datadistributor/DataDistribution.cpp | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/fdbserver/datadistributor/DataDistribution.cpp b/fdbserver/datadistributor/DataDistribution.cpp
index f6aa29b94e9..b8e22b0d43e 100644
--- a/fdbserver/datadistributor/DataDistribution.cpp
+++ b/fdbserver/datadistributor/DataDistribution.cpp
@@ -2912,8 +2912,7 @@ Future<Void> dataDistribution(Reference<DataDistributor> self,
 				                                // reach ACCEPTING_COMMITS before continuing startup. Waiting for
 				                                // ALL_LOGS_RECRUITED can hang during degraded failover scenarios
 				                                // such as when the primary DC is down.
-				                                self->initialized.getFuture() &&
-				                                    waitForAcceptingCommits(self->dbInfo),
+				                                self->initialized.getFuture() && waitForAcceptingCommits(self->dbInfo),
 				                                zeroHealthyTeams[1],
 				                                IsPrimary::False,
 				                                processingUnhealthy,