[tarantool-patches] [PATCH 1/1] replication: recfg with 0 quorum returns immediately

[Vladislav Shpilevoy]

Replication quorum 0 not only affects orphan status, but also,
according to documentation, makes box.cfg() return immediately
regardless of whether connections to upstreams are established.

It was not so before the patch. What is worse, even with non 0
quorum the instance was blocked on reconfiguration for connect
timeout seconds, if at least one node is not connected.

Now quorum is respected on reconfiguration. On a bootstrap it is
still impossible to return earlier than
replication_connect_timeout, because nodes need to choose some
cluster settings. Too early start would make it impossible -
cluster's participants will just start and choose different
cluster UUIDs.

Closes #3760
---
Branch: https://github.com/tarantool/tarantool/tree/gerold103/gh-3760-zero-quorum-recfg
Issue: https://github.com/tarantool/tarantool/issues/3760

 src/box/replication.cc         | 14 ++++++++++++++
 test/replication/misc.result   | 27 ++++++++++++++++++++++++---
 test/replication/misc.test.lua | 24 +++++++++++++++++++-----
 3 files changed, 57 insertions(+), 8 deletions(-)

diff --git a/src/box/replication.cc b/src/box/replication.cc
index d691ce487..ea88daf77 100644
--- a/src/box/replication.cc
+++ b/src/box/replication.cc
@@ -659,6 +659,20 @@ replicaset_connect(struct applier **appliers, int count,
 	}
 
 	while (state.connected < count) {
+		/*
+		 * After a quorum is reached, it is considered
+		 * enough to proceed. Except if a connection is
+		 * critical.
+		 * Connection *is* critical even with 0 quorum
+		 * when the instance starts first time and needs
+		 * to choose replicaset UUID, fill _cluster, etc.
+		 * If 0 quorum allowed to return immediately even
+		 * at first start, then it would be impossible to
+		 * bootstrap a replicaset - all nodes would start
+		 * immediately and choose different cluster UUIDs.
+		 */
+		if (state.connected >= quorum && !connect_quorum)
+			break;
 		double wait_start = ev_monotonic_now(loop());
 		if (fiber_cond_wait_timeout(&state.wakeup, timeout) != 0)
 			break;
diff --git a/test/replication/misc.result b/test/replication/misc.result
index ae72ce3e4..3905cc49e 100644
--- a/test/replication/misc.result
+++ b/test/replication/misc.result
@@ -818,12 +818,33 @@ box.info.ro
 ---
 - false
 ...
-box.cfg{replication=""}
+--
+-- gh-3760: replication quorum 0 on reconfiguration should return
+-- from box.cfg immediately.
+--
+replication = box.cfg.replication
+---
+...
+box.cfg{                                                        \
+    replication = {},                                           \
+    replication_connect_quorum = 0,                             \
+    replication_connect_timeout = 1000000                       \
+}
 ---
 ...
-box.cfg{replication_connect_timeout=replication_connect_timeout}
+-- The call below would hang, if quorum 0 is ignored, or checked
+-- too late.
+box.cfg{replication = {'localhost:12345'}}
 ---
 ...
-box.cfg{replication_connect_quorum=replication_connect_quorum}
+box.info.status
+---
+- running
+...
+box.cfg{                                                        \
+    replication = replication,                                  \
+    replication_connect_quorum = replication_connect_quorum,    \
+    replication_connect_timeout = replication_connect_timeout   \
+}
 ---
 ...
diff --git a/test/replication/misc.test.lua b/test/replication/misc.test.lua
index 16e7e9e42..696564f94 100644
--- a/test/replication/misc.test.lua
+++ b/test/replication/misc.test.lua
@@ -327,8 +327,22 @@ box.cfg{replication_connect_quorum=1}
 box.info.status
 box.info.ro
 
-box.cfg{replication=""}
-
-
-box.cfg{replication_connect_timeout=replication_connect_timeout}
-box.cfg{replication_connect_quorum=replication_connect_quorum}
+--
+-- gh-3760: replication quorum 0 on reconfiguration should return
+-- from box.cfg immediately.
+--
+replication = box.cfg.replication
+box.cfg{                                                        \
+    replication = {},                                           \
+    replication_connect_quorum = 0,                             \
+    replication_connect_timeout = 1000000                       \
+}
+-- The call below would hang, if quorum 0 is ignored, or checked
+-- too late.
+box.cfg{replication = {'localhost:12345'}}
+box.info.status
+box.cfg{                                                        \
+    replication = replication,                                  \
+    replication_connect_quorum = replication_connect_quorum,    \
+    replication_connect_timeout = replication_connect_timeout   \
+}
-- 
2.21.0 (Apple Git-122)