[PATCH] vinyl: cancel reader and writer threads on shutdown

[Vladimir Davydov <vdavydov.dev@gmail.com>]

Currently, vinyl won't shutdown until all reader and writer threads
gracefully complete all their pending requests, which may take a while,
especially for writer threads that may happen to be doing compaction at
the time. This is annoying - there's absolutely no reason to delay
termination in such a case. Let's forcefully cancel all threads, like we
do in case of relay threads.

This should fix sporadic vinyl/replica_quota test hang.

Closes #3949
---
https://github.com/tarantool/tarantool/issues/3949
https://github.com/tarantool/tarantool/commits/dv/gh-3949-vy-cancel-threads-on-shutdown

 src/box/vy_run.c           |  7 ++-----
 src/box/vy_scheduler.c     |  5 ++---
 test/vinyl/errinj.result   | 40 ++++++++++++++++++++++++++++++++++++++++
 test/vinyl/errinj.test.lua | 20 ++++++++++++++++++++
 4 files changed, 64 insertions(+), 8 deletions(-)

diff --git a/src/box/vy_run.c b/src/box/vy_run.c
index cee90458..9aa3beec 100644
--- a/src/box/vy_run.c
+++ b/src/box/vy_run.c
@@ -152,11 +152,8 @@ vy_run_env_stop_readers(struct vy_run_env *env)
 {
 	for (int i = 0; i < env->reader_pool_size; i++) {
 		struct vy_run_reader *reader = &env->reader_pool[i];
-
-		cbus_stop_loop(&reader->reader_pipe);
-		cpipe_destroy(&reader->reader_pipe);
-		if (cord_join(&reader->cord) != 0)
-			panic("failed to join vinyl reader thread");
+		tt_pthread_cancel(reader->cord.id);
+		tt_pthread_join(reader->cord.id, NULL);
 	}
 	free(env->reader_pool);
 }
diff --git a/src/box/vy_scheduler.c b/src/box/vy_scheduler.c
index 5ec6d171..d4047197 100644
--- a/src/box/vy_scheduler.c
+++ b/src/box/vy_scheduler.c
@@ -371,9 +371,8 @@ vy_worker_pool_stop(struct vy_worker_pool *pool)
 	assert(pool->workers != NULL);
 	for (int i = 0; i < pool->size; i++) {
 		struct vy_worker *worker = &pool->workers[i];
-		cbus_stop_loop(&worker->worker_pipe);
-		cpipe_destroy(&worker->worker_pipe);
-		cord_join(&worker->cord);
+		tt_pthread_cancel(worker->cord.id);
+		tt_pthread_join(worker->cord.id, NULL);
 	}
 	free(pool->workers);
 	pool->workers = NULL;
diff --git a/test/vinyl/errinj.result b/test/vinyl/errinj.result
index 990c7e85..248b32c8 100644
--- a/test/vinyl/errinj.result
+++ b/test/vinyl/errinj.result
@@ -1126,3 +1126,43 @@ box.schema.user.revoke('guest', 'replication')
 s:drop()
 ---
 ...
+--
+-- Check that tarantool stops immediately even if a vinyl worker
+-- thread is blocked (see gh-3225).
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+---
+...
+_ = s:create_index('pk')
+---
+...
+s:replace{1, 1}
+---
+- [1, 1]
+...
+box.snapshot()
+---
+- ok
+...
+errinj.set('ERRINJ_VY_READ_PAGE_TIMEOUT', 9000)
+---
+- ok
+...
+_ = fiber.create(function() s:get(1) end)
+---
+...
+s:replace{1, 2}
+---
+- [1, 2]
+...
+errinj.set('ERRINJ_VY_RUN_WRITE_STMT_TIMEOUT', 9000)
+---
+- ok
+...
+_ = fiber.create(function() box.snapshot() end)
+---
+...
+test_run:cmd("restart server default")
+box.space.test:drop()
+---
+...
diff --git a/test/vinyl/errinj.test.lua b/test/vinyl/errinj.test.lua
index d374a910..eaec52a5 100644
--- a/test/vinyl/errinj.test.lua
+++ b/test/vinyl/errinj.test.lua
@@ -408,3 +408,23 @@ test_run:cmd("delete server replica")
 errinj.set('ERRINJ_VYRUN_INDEX_GARBAGE', false)
 box.schema.user.revoke('guest', 'replication')
 s:drop()
+
+--
+-- Check that tarantool stops immediately even if a vinyl worker
+-- thread is blocked (see gh-3225).
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+_ = s:create_index('pk')
+s:replace{1, 1}
+box.snapshot()
+
+errinj.set('ERRINJ_VY_READ_PAGE_TIMEOUT', 9000)
+_ = fiber.create(function() s:get(1) end)
+
+s:replace{1, 2}
+
+errinj.set('ERRINJ_VY_RUN_WRITE_STMT_TIMEOUT', 9000)
+_ = fiber.create(function() box.snapshot() end)
+
+test_run:cmd("restart server default")
+box.space.test:drop()
-- 
2.11.0