From 7d093bb19cbc1409441286bacb35523ec25b1d72 Mon Sep 17 00:00:00 2001
From: Vladislav Shpilevoy <v.shpilevoy@tarantool.org>
Date: Mon, 27 Feb 2023 23:08:34 +0100
Subject: [PATCH] test: add a test for join with self delete
Deletion of the own entry from _cluster space is allowed during
the join stage, because the remote master could have already had
the joining instance UUID in _cluster space but then deleted it.
Then for the joining instance it looks like deletion of self from
_cluster. But that is fine - in the end of join the master will
register the replica again.
The case is handled, but not covered with a test. The patch adds
one.
NO_DOC=test
NO_CHANGELOG=test
---
test/replication-luatest/join_test.lua | 66 ++++++++++++++++++++++++++
1 file changed, 66 insertions(+)
create mode 100644 test/replication-luatest/join_test.lua
diff --git a/test/replication-luatest/join_test.lua b/test/replication-luatest/join_test.lua
new file mode 100644
index 0000000000..75648efe29
--- /dev/null
+++ b/test/replication-luatest/join_test.lua
@@ -0,0 +1,66 @@
+local t = require('luatest')
+local server = require('luatest.server')
+local uuid = require('uuid')
+
+local g = t.group('join')
+
+g.before_all(function(lg)
+ lg.master = server:new({
+ alias = 'master',
+ box_cfg = {
+ replication_timeout = 0.1,
+ },
+ })
+ lg.master:start()
+end)
+
+g.after_all(function(lg)
+ lg.master:drop()
+end)
+
+--
+-- New replica (or one being re-bootstrapped) can legally see deletion of self
+-- from _cluster space during join. It is allowed because the master could
+-- delete the replica from _cluster before the replica tried to rejoin or join
+-- first time via a previously existing UUID.
+--
+g.test_fetch_self_delete_during_final_join = function(lg)
+ t.tarantool.skip_if_not_debug()
+ lg.master:exec(function()
+ local s = box.schema.create_space('test')
+ s:create_index('pk')
+ s:replace{1}
+ box.error.injection.set('ERRINJ_RELAY_SEND_DELAY', true)
+ end)
+ local replica_uuid = uuid.str()
+ local box_cfg = table.deepcopy(lg.master.box_cfg)
+ box_cfg.replication = {lg.master.net_box_uri}
+ box_cfg.instance_uuid = replica_uuid
+ local replica = server:new({
+ alias = 'replica',
+ box_cfg = box_cfg,
+ })
+ replica:start({wait_until_ready = false})
+ local msg = ('joining replica %s'):format(replica_uuid):gsub('%-', '%%-')
+ t.helpers.retrying({}, function()
+ t.assert(lg.master:grep_log(msg))
+ end)
+ lg.master:exec(function(replica_uuid)
+ local _cluster = box.space._cluster
+ _cluster:replace{2, replica_uuid}
+ _cluster:delete{2}
+ _cluster:replace{2, replica_uuid}
+ _cluster:delete{2}
+ box.error.injection.set('ERRINJ_RELAY_SEND_DELAY', false)
+ end, {replica_uuid})
+ replica:wait_until_ready()
+ local replica_id = replica:exec(function()
+ t.assert_equals(box.space.test:get{1}, {1})
+ return box.info.id
+ end)
+ replica:drop()
+ lg.master:exec(function(replica_id)
+ box.space.test:drop()
+ box.space._cluster:delete{replica_id}
+ end, {replica_id})
+end
--
GitLab