diff --git a/src/box/box.cc b/src/box/box.cc
index cc0b8ed739faed6829aeeccceda25bb129b639f7..ed2627771dce06cb68eb2a7ad51351ae1dcfb950 100644
--- a/src/box/box.cc
+++ b/src/box/box.cc
@@ -271,6 +271,15 @@ box_check_writable(void)
error_set_uuid(e, "queue_owner_uuid", &r->uuid);
error_append_msg(e, " (%s)", tt_uuid_str(&r->uuid));
}
+ if (txn_limbo.owner_id == instance_id) {
+ if (txn_limbo.is_frozen_due_to_fencing) {
+ error_append_msg(e, " and is frozen due to "
+ "fencing");
+ } else if (txn_limbo.is_frozen_until_promotion) {
+ error_append_msg(e, " and is frozen until "
+ "promotion");
+ }
+ }
} else {
if (is_ro)
error_append_msg(e, "box.cfg.read_only is true");
@@ -2005,8 +2014,10 @@ box_promote(void)
* Currently active leader (the instance that is seen as leader by both
* raft and txn_limbo) can't issue another PROMOTE.
*/
- bool is_leader = txn_limbo_replica_term(&txn_limbo, instance_id) ==
- raft->term && txn_limbo.owner_id == instance_id;
+ bool is_leader =
+ txn_limbo_replica_term(&txn_limbo, instance_id) == raft->term &&
+ txn_limbo.owner_id == instance_id &&
+ !txn_limbo.is_frozen_until_promotion;
if (box_election_mode != ELECTION_MODE_OFF)
is_leader = is_leader && raft->state == RAFT_STATE_LEADER;
diff --git a/src/box/txn_limbo.c b/src/box/txn_limbo.c
index d85e33bb2ca3db73848c56177de099c30607c9f6..f8382243cb63be7820c7f0e31087955c1dd325e8 100644
--- a/src/box/txn_limbo.c
+++ b/src/box/txn_limbo.c
@@ -54,6 +54,7 @@ txn_limbo_create(struct txn_limbo *limbo)
limbo->is_in_rollback = false;
limbo->svp_confirmed_lsn = -1;
limbo->frozen_reasons = 0;
+ limbo->is_frozen_until_promotion = true;
}
static inline bool
@@ -830,6 +831,16 @@ txn_limbo_req_rollback(struct txn_limbo *limbo,
}
}
+/** Unfreeze the limbo encountering the first new PROMOTE after a restart. */
+static inline void
+txn_limbo_unfreeze_on_first_promote(struct txn_limbo *limbo)
+{
+ if (box_is_configured()) {
+ limbo->is_frozen_until_promotion = false;
+ box_update_ro_summary();
+ }
+}
+
void
txn_limbo_req_commit(struct txn_limbo *limbo, const struct synchro_request *req)
{
@@ -854,9 +865,11 @@ txn_limbo_req_commit(struct txn_limbo *limbo, const struct synchro_request *req)
vclock_follow(&limbo->promote_term_map, origin, term);
if (term > limbo->promote_greatest_term) {
limbo->promote_greatest_term = term;
- if (iproto_type_is_promote_request(req->type) &&
- term >= box_raft()->volatile_term)
- txn_limbo_unfence(&txn_limbo);
+ if (iproto_type_is_promote_request(req->type)) {
+ if (term >= box_raft()->volatile_term)
+ txn_limbo_unfence(limbo);
+ txn_limbo_unfreeze_on_first_promote(&txn_limbo);
+ }
}
} else if (iproto_type_is_promote_request(req->type) &&
limbo->promote_greatest_term > 1) {
diff --git a/src/box/txn_limbo.h b/src/box/txn_limbo.h
index 095c965a0c53c9ec075682ae3df5ef6644668b66..7167065c221c664bc96a13f366948324a3d284df 100644
--- a/src/box/txn_limbo.h
+++ b/src/box/txn_limbo.h
@@ -207,6 +207,12 @@ struct txn_limbo {
* remote instance.
*/
bool is_frozen_due_to_fencing : 1;
+ /*
+ * This mode is always on upon node start and is turned
+ * off by any new PROMOTE arriving either via
+ * replication or issued by the node.
+ */
+ bool is_frozen_until_promotion : 1;
};
};
};
diff --git a/test/replication/gh-5140-qsync-casc-rollback.result b/test/replication/gh-5140-qsync-casc-rollback.result
index d3208e1a4f210b87ddbf221422e6e512d6bb6c4a..a71d8f9818cd40f8a9a61ed42c9a782b87899840 100644
--- a/test/replication/gh-5140-qsync-casc-rollback.result
+++ b/test/replication/gh-5140-qsync-casc-rollback.result
@@ -206,6 +206,10 @@ box.space.sync:select{}
| - [4]
| ...
+box.ctl.promote()
+ | ---
+ | ...
+
box.space.sync:drop()
| ---
| ...
diff --git a/test/replication/gh-5140-qsync-casc-rollback.test.lua b/test/replication/gh-5140-qsync-casc-rollback.test.lua
index 96ddfd2609dbff795443a2d155bb03c0a7f54436..15732e332414da390dd7707513b52b33f2b0bf11 100644
--- a/test/replication/gh-5140-qsync-casc-rollback.test.lua
+++ b/test/replication/gh-5140-qsync-casc-rollback.test.lua
@@ -97,6 +97,8 @@ test_run:switch('default')
box.space.async:select{}
box.space.sync:select{}
+box.ctl.promote()
+
box.space.sync:drop()
box.space.async:drop()
diff --git a/test/replication/gh-5163-qsync-restart-crash.result b/test/replication/gh-5163-qsync-restart-crash.result
index 1b4d3d9b5cc1baaad43c3ae6547d93671d99929b..79bf96e474ecc2d8350af7e12d4d5b94e332ae5d 100644
--- a/test/replication/gh-5163-qsync-restart-crash.result
+++ b/test/replication/gh-5163-qsync-restart-crash.result
@@ -30,6 +30,9 @@ box.space.sync:select{}
| ---
| - - [1]
| ...
+box.ctl.promote()
+ | ---
+ | ...
box.space.sync:drop()
| ---
| ...
diff --git a/test/replication/gh-5163-qsync-restart-crash.test.lua b/test/replication/gh-5163-qsync-restart-crash.test.lua
index c8d54aad2397e37d30df72a693e88f20def9deda..0298d8ce6e30e232bfca2eb0c79a5dc63fcf5d56 100644
--- a/test/replication/gh-5163-qsync-restart-crash.test.lua
+++ b/test/replication/gh-5163-qsync-restart-crash.test.lua
@@ -12,5 +12,6 @@ box.ctl.promote()
box.space.sync:replace{1}
test_run:cmd('restart server default')
box.space.sync:select{}
+box.ctl.promote()
box.space.sync:drop()
box.ctl.demote()
diff --git a/test/replication/gh-5288-qsync-recovery.result b/test/replication/gh-5288-qsync-recovery.result
index 704b71d930255f3eb5e5e3116096a71adcd3a27e..dc796181d57e2ebf1efe82453ec4c3b8c707568b 100644
--- a/test/replication/gh-5288-qsync-recovery.result
+++ b/test/replication/gh-5288-qsync-recovery.result
@@ -25,6 +25,9 @@ box.snapshot()
| ...
test_run:cmd('restart server default')
|
+box.ctl.promote()
+ | ---
+ | ...
box.space.sync:drop()
| ---
| ...
diff --git a/test/replication/gh-5288-qsync-recovery.test.lua b/test/replication/gh-5288-qsync-recovery.test.lua
index 2455f7278abe28b9e84d14c21140ac098ead084c..095bc71f9b95166ef75c7ce848ad47ba6a163d31 100644
--- a/test/replication/gh-5288-qsync-recovery.test.lua
+++ b/test/replication/gh-5288-qsync-recovery.test.lua
@@ -9,5 +9,6 @@ box.ctl.promote()
s:insert{1}
box.snapshot()
test_run:cmd('restart server default')
+box.ctl.promote()
box.space.sync:drop()
box.ctl.demote()
diff --git a/test/replication/gh-5298-qsync-recovery-snap.result b/test/replication/gh-5298-qsync-recovery-snap.result
index 0883fe5f5e1a5b66c46afa4aa05a8b6945f00c2d..52e13d75d07457ce9bc269d2d37c4018de16b283 100644
--- a/test/replication/gh-5298-qsync-recovery-snap.result
+++ b/test/replication/gh-5298-qsync-recovery-snap.result
@@ -43,58 +43,16 @@ box.snapshot()
test_run:cmd("restart server default")
|
--- Could hang if the limbo would incorrectly handle the snapshot end.
-box.space.sync:replace{11}
+-- Would be non-empty if limbo would incorrectly handle the snapshot end.
+box.info.synchro.queue.len
| ---
- | - [11]
+ | - 0
| ...
-old_synchro_quorum = box.cfg.replication_synchro_quorum
- | ---
- | ...
-old_synchro_timeout = box.cfg.replication_synchro_timeout
- | ---
- | ...
-
-box.cfg{ \
- replication_synchro_timeout = 0.001, \
- replication_synchro_quorum = 2, \
-}
- | ---
- | ...
-box.space.sync:replace{12}
- | ---
- | - error: Quorum collection for a synchronous transaction is timed out
- | ...
-
-box.cfg{ \
- replication_synchro_timeout = 1000, \
- replication_synchro_quorum = 1, \
-}
- | ---
- | ...
-box.space.sync:replace{13}
- | ---
- | - [13]
- | ...
-box.space.sync:get({11})
- | ---
- | - [11]
- | ...
-box.space.sync:get({12})
- | ---
- | ...
-box.space.sync:get({13})
+box.ctl.promote()
| ---
- | - [13]
| ...
-box.cfg{ \
- replication_synchro_timeout = old_synchro_timeout, \
- replication_synchro_quorum = old_synchro_quorum, \
-}
- | ---
- | ...
box.space.sync:drop()
| ---
| ...
diff --git a/test/replication/gh-5298-qsync-recovery-snap.test.lua b/test/replication/gh-5298-qsync-recovery-snap.test.lua
index 084cde963d8ef087cddf69d855e51248c06f8aea..55d9501647edce87aaef039ee65eb8eadb3d760b 100644
--- a/test/replication/gh-5298-qsync-recovery-snap.test.lua
+++ b/test/replication/gh-5298-qsync-recovery-snap.test.lua
@@ -20,31 +20,11 @@ box.snapshot()
test_run:cmd("restart server default")
--- Could hang if the limbo would incorrectly handle the snapshot end.
-box.space.sync:replace{11}
+-- Would be non-empty if limbo would incorrectly handle the snapshot end.
+box.info.synchro.queue.len
-old_synchro_quorum = box.cfg.replication_synchro_quorum
-old_synchro_timeout = box.cfg.replication_synchro_timeout
-
-box.cfg{ \
- replication_synchro_timeout = 0.001, \
- replication_synchro_quorum = 2, \
-}
-box.space.sync:replace{12}
-
-box.cfg{ \
- replication_synchro_timeout = 1000, \
- replication_synchro_quorum = 1, \
-}
-box.space.sync:replace{13}
-box.space.sync:get({11})
-box.space.sync:get({12})
-box.space.sync:get({13})
+box.ctl.promote()
-box.cfg{ \
- replication_synchro_timeout = old_synchro_timeout, \
- replication_synchro_quorum = old_synchro_quorum, \
-}
box.space.sync:drop()
box.space.loc:drop()
box.ctl.demote()
diff --git a/test/replication/gh-5874-qsync-txn-recovery.result b/test/replication/gh-5874-qsync-txn-recovery.result
index 01328a9e32cfda88bba221cbe6525e23426397a5..11f2fc9c79358b447c459619c16d8981c26175e5 100644
--- a/test/replication/gh-5874-qsync-txn-recovery.result
+++ b/test/replication/gh-5874-qsync-txn-recovery.result
@@ -154,6 +154,10 @@ loc:select()
| - [2]
| - [3]
| ...
+
+box.ctl.promote()
+ | ---
+ | ...
async:drop()
| ---
| ...
diff --git a/test/replication/gh-5874-qsync-txn-recovery.test.lua b/test/replication/gh-5874-qsync-txn-recovery.test.lua
index 6ddf164ace1af74cc286c48f8c6dd5c7c945cc62..5bf3ba206ef7d9345033ca22774105af46f502f4 100644
--- a/test/replication/gh-5874-qsync-txn-recovery.test.lua
+++ b/test/replication/gh-5874-qsync-txn-recovery.test.lua
@@ -80,6 +80,8 @@ loc = box.space.loc
async:select()
sync:select()
loc:select()
+
+box.ctl.promote()
async:drop()
sync:drop()
loc:drop()
diff --git a/test/replication/suite.cfg b/test/replication/suite.cfg
index 3eee0803c5dae9c06fd27a6692ba31c554e56b19..8ed9eeb0343b8c91457df1fd1fcacc2b93bd2295 100644
--- a/test/replication/suite.cfg
+++ b/test/replication/suite.cfg
@@ -18,6 +18,7 @@
"gh-4424-misc-orphan-on-reconfiguration-error.test.lua": {},
"gh-5213-qsync-applier-order.test.lua": {},
"gh-5213-qsync-applier-order-3.test.lua": {},
+ "gh-5288-qsync-recovery.test.lua": {},
"gh-5426-election-on-off.test.lua": {},
"gh-5430-cluster-mvcc.test.lua": {},
"gh-5433-election-restart-recovery.test.lua": {},