Debian Codegraph @

mysql-test/suite/group_replication/t/gr_majority_loss_2_to_1.test

raw
1 ################################################################################
2 # This test validates that in a group with two members, if one of the members
3 # goes away, the group will block. In addition, it also checks that if the
4 # group is forcefully reconfigured the system will resume operations.
5 #
6 # It is a variation of other reconfiguration tests that we have deployed
7 # already, except that this tests the case of 2 servers only.
8 #
9 # The test does the following:
10 # 0. The test requires two servers: server1 and server2.
11 # 1. Bootstrap start group on server1. Start GR on server2. Create table on
12 # server2 for replicating data.
13 # 2. Crashes one server: server2.
14 # 3. Issue a transaction on the server1 that is still up and validate that it
15 # blocks (best effort test).
16 # 4. Validate that the group configuration still shows up as two servers.
17 # 5. Unblock the group by reconfiguring it forcefully to a single server.
18 # 6. Check that the pending transaction was unblocked.
19 # 7. Validate that the configuration is now a single server.
20 # 8. Issue a new transaction and validate that it completes.
21 # 9. Join server2 again.
22 # 10. Check that the group configuration has changed on both servers.
23 # 11. Check that the contents of the replicated table are in sync.
24 ################################################################################
25 # This test does crashes servers, thence we skip it on valgrind.
26 --source include/not_valgrind.inc
27 --source include/big_test.inc
28 --source include/force_restart.inc
29 --source include/have_group_replication_plugin.inc
30
31 ###
32 ### 1. sets up the group with two servers: server1 and server2
33 ###
34
35 --let $rpl_server_count= 2
36 --let $rpl_skip_group_replication_start= 1
37 --source include/group_replication.inc
38
39 --let $rpl_connection_name= server1
40 --source include/rpl_connection.inc
41 --source include/start_and_bootstrap_group_replication.inc
42 --let $member1_uuid= query_get_value(SELECT @@SERVER_UUID, @@SERVER_UUID, 1)
43
44 --let $rpl_connection_name= server2
45 --source include/rpl_connection.inc
46 --let $member2_uuid= query_get_value(SELECT @@SERVER_UUID, @@SERVER_UUID, 1)
47
48 SET SQL_LOG_BIN=0;
49 call mtr.add_suppression("site definition is null");
50 SET SQL_LOG_BIN=1;
51
52 # start the server
53 --source include/start_group_replication.inc
54
55 # save this, since it will be used to recover the server later on
56 --let $local_address_server2= `SELECT @@GLOBAL.group_replication_local_address`
57 --let $group_seeds_server2= `SELECT @@GLOBAL.group_replication_group_seeds`
58
59 # create a table for replicating data
60 CREATE TABLE t1 (c1 INT NOT NULL PRIMARY KEY) ENGINE=InnoDB;
61 --source include/rpl_sync.inc
62
63 ###
64 ### 2. crashes one server: server2
65 ###
66
67 #
68 # We do kill and restart the servers. We wait for the server to start.
69 #
70 # However, since the servers loose its GCS configuration they
71 # will not rejoin the group.
72 #
73 --let $rpl_connection_name= server2
74 --source include/rpl_connection.inc
75
76 --source include/kill_and_restart_mysqld.inc
77
78 --let $rpl_connection_name= server_1
79 --source include/rpl_connection.inc
80
81 #
82 # Wait for the member state be unreachable, meaning
83 # that we have lost quorum and we are conveying the
84 # information to the user that the we are unable to
85 # speak to server 2.
86 #
87 --let $group_replication_member_state= UNREACHABLE
88 --let $group_replication_member_id= $member2_uuid
89 --source include/gr_wait_for_member_state.inc
90 # reset the group_replication_id var
91 --let $group_replication_member_id= $member1_uuid
92
93 --let $rpl_server_number= 2
94 --source include/rpl_reconnect.inc
95
96 --let $rpl_connection_name= server_2
97 --source include/rpl_connection.inc
98
99 --replace_result $group_seeds_server2 GROUP_SEEDS_SERVER2
100 --eval SET @@global.group_replication_group_seeds="$group_seeds_server2"
101 --replace_result $local_address_server2 LOCAL_ADDRESS_SERVER2
102 --eval SET @@global.group_replication_local_address="$local_address_server2"
103 --replace_result $group_replication_group_name GROUP_REPLICATION_GROUP_NAME
104 --eval SET @@global.group_replication_group_name="$group_replication_group_name"
105
106 ###
107 ### 3. issue a transaction on the server that is still up
108 ### and validate that it blocks (best effort test)
109 ###
110 # Group is blocked and so cannot change its configuration.
111 --let $rpl_connection_name= server_1
112 --source include/rpl_connection.inc
113 --let $group_replication_number_of_members= 2
114 --source include/gr_wait_for_number_of_members.inc
115
116 # This transaction will be blocked since group lost its majority.
117 # Though it needs to be delivered to the group when unblocked.
118 --send INSERT INTO t1 VALUES (1)
119
120 --let $rpl_connection_name= server1
121 --source include/rpl_connection.inc
122 --let $local_address_server1= `SELECT @@GLOBAL.group_replication_local_address`
123
124 # wait until the transaction blocks on "waiting for handler commit"
125 --let $wait_condition= select PROCESSLIST_STATE="waiting for handler commit" from performance_schema.threads where processlist_INFO LIKE "INSERT INTO t1 VALUES (1)"
126 --source include/wait_condition.inc
127
128 ###
129 ### 4. validate that the group configuration still shows
130 ### up as two servers (and one is unreachable)
131 ###
132
133 # verify that there are still two members in the group
134 --let $group_replication_number_of_members= 2
135 --source include/gr_wait_for_number_of_members.inc
136
137 --let $assert_text= t1 does not have 2 inserted into it yet
138 --let $assert_cond= COUNT(*)=0 FROM t1
139 --source include/assert.inc
140
141 --let $res= `SELECT COUNT(*)=1 FROM performance_schema.replication_group_members WHERE member_id="$member2_uuid" AND member_state="UNREACHABLE"`
142 --let $assert_text= server 2 is marked as unreachable
143 --let $assert_cond= $res = 1
144 --source include/assert.inc
145
146 ###
147 ### 5. Unblock the group by reconfiguring it forcefully to a
148 ### single server.
149
150 --disable_query_log
151 --eval SET GLOBAL group_replication_force_members= "$local_address_server1"
152 --enable_query_log
153
154 --let $assert_text= group_replication_force_members is correct
155 --let $assert_cond= "[SELECT @@GLOBAL.group_replication_force_members]" = "$local_address_server1"
156 --source include/assert.inc
157
158 # verify that the table has now only 1 member
159 --let $group_replication_number_of_members= 1
160 --source include/gr_wait_for_number_of_members.inc
161
162 # no more unreachable entries in the table
163 --let $res= `SELECT COUNT(*)=0 FROM performance_schema.replication_group_members WHERE member_state="UNREACHABLE"`
164 --let $assert_text= No more unreachable servers in the P_S table.
165 --let $assert_cond= $res = 1
166 --source include/assert.inc
167
168 ###
169 ### 6. Check that the pending transaction was unblocked.
170 ###
171
172 # the transaction will resume now
173 --let $wait_condition= SELECT COUNT(*)=1 FROM t1
174 --source include/wait_condition.inc
175
176 # reap the transaction that had been waiting for the majority
177 --let $rpl_connection_name= server_1
178 --source include/rpl_connection.inc
179 --reap
180
181 ###
182 ### 7. validate that the configuration is now a single server
183 ###
184
185 --let $rpl_connection_name= server1
186 --source include/rpl_connection.inc
187
188 --let $group_replication_number_of_members= 1
189 --source include/gr_wait_for_number_of_members.inc
190
191 --let $server_uuid= `SELECT @@GLOBAL.SERVER_UUID`
192 --let $assert_text= Server 1 must be present on group members
193 --let $assert_cond= [SELECT COUNT(*) FROM performance_schema.replication_group_members WHERE member_id="$server_uuid"] = 1
194 --source include/assert.inc
195
196 --let $raw_view_id= query_get_value(SELECT view_id FROM performance_schema.replication_group_member_stats, view_id, 1)
197 --let $view_id= `SELECT RIGHT('$raw_view_id', 1)`
198 --let $assert_text= The value of view_id must be random:3
199 --let $assert_cond= $view_id >= 3
200 --source include/assert.inc
201
202 ###
203 ### 8. issue a new transaction and validate that it completes.
204 ###
205 --let $rpl_connection_name= server1
206 --source include/rpl_connection.inc
207 INSERT INTO t1 VALUES (2);
208
209 --let $assert_text= 'There are two values in table t1'
210 --let $assert_cond= [SELECT COUNT(*) AS count FROM t1, count, 1] = 2
211 --source include/assert.inc
212
213 ###
214 ### 9. join server 2 again
215 ###
216
217 --let $rpl_connection_name= server2
218 --source include/rpl_connection.inc
219
220 # this already waits for recovery to happen
221 --source include/start_group_replication.inc
222
223 --let $rpl_connection_name= server1
224 --source include/rpl_connection.inc
225
226 # sync just to make sure
227 --source include/rpl_sync.inc
228
229 ###
230 ### 10. check that the group configuration has changed on both
231 ### servers
232
233 # validate that there are two servers in the group
234 --let $group_replication_number_of_members= 2
235 --source include/gr_wait_for_number_of_members.inc
236
237 --let $rpl_connection_name= server2
238 --source include/rpl_connection.inc
239
240 # validate that there are two servers in the group
241 --let $group_replication_number_of_members= 2
242 --source include/gr_wait_for_number_of_members.inc
243
244 ###
245 ### 11. check that the contents of the replicated table are
246 ### in sync
247 ###
248
249 # validate that the table content is in sync
250 --let diff_tables=server1:test.t1,server2:test.t1
251 --source include/diff_tables.inc
252
253 ### Cleanup
254
255 --let $rpl_connection_name= server1
256 --source include/rpl_connection.inc
257
258 DROP table t1;
259
260 --source include/group_replication_end.inc