Debian Codegraph @

mysql-test/suite/innodb/r/log_flush_order.result

raw
1 #
2 # BUG#27664539 INNODB: ASSERTION FAILURE: BUF0FLU.CC:457:
3 # (BUF_POOL->FLUSH_LIST).START == __NULL
4 #
5 #
6 # Scenario 1.1.1 - No-redo dirtied page inserted too quickly.
7 # Non-empty flush list when it is inserted.
8 # The page is no-redo dirtied first time.
9 #
10 #
11 # 0. Block log.recent_closed.tail() from advancing by stopping single mtr
12 # just before it inserts its dirty pages to flush list.
13 #
14 #
15 # 1. Optionally modify no-redo dirtied page (then it is later remodified).
16 #
17 #
18 # 2. Gather threads that try to insert dirtied pages with underlying redo
19 # records that protect the changes.
20 #
21 #
22 # 3. Wait gathering such threads until more data is written to log buffer.
23 # Wait until current lsn is greater than log.recent_closed.tail() by
24 # more than capacity of log.recent_closed.
25 #
26 include/assert.inc [We must advance current lsn by more than maximum flush order lag.]
27 #
28 # 4. Execute mini-transactions with no-redo dirtied pages (MTR_LOG_NO_REDO).
29 #
30 # When bug was present, such mtr inserted pages with too big lsn using
31 # current lsn, which does not fit space in log.recent_closed, bypassing
32 # required wait for space there (thanks to MTR_LOG_NO_REDO).
33 #
34 #
35 # 5. Resume the old mtr which blocked the log.recent_closed.tail() from advancing.
36 #
37 #
38 # 6. Disconnect pending connections.
39 #
40 #
41 # 7. Perform standard mtrs, just to ensure all is fine.
42 #
43 include/assert.inc [We must not have oldest_lsn_approx < checkpoint_lsn.]
44 # Kill and restart
45 #
46 # Scenario 1.1.2 - No-redo dirtied page inserted too quickly.
47 # Non-empty flush list when it is inserted.
48 # The page is no-redo dirtied second time.
49 #
50 #
51 # 0. Block log.recent_closed.tail() from advancing by stopping single mtr
52 # just before it inserts its dirty pages to flush list.
53 #
54 #
55 # 1. Optionally modify no-redo dirtied page (then it is later remodified).
56 #
57 #
58 # 2. Gather threads that try to insert dirtied pages with underlying redo
59 # records that protect the changes.
60 #
61 #
62 # 3. Wait gathering such threads until more data is written to log buffer.
63 # Wait until current lsn is greater than log.recent_closed.tail() by
64 # more than capacity of log.recent_closed.
65 #
66 include/assert.inc [We must advance current lsn by more than maximum flush order lag.]
67 #
68 # 4. Execute mini-transactions with no-redo dirtied pages (MTR_LOG_NO_REDO).
69 #
70 # When bug was present, such mtr inserted pages with too big lsn using
71 # current lsn, which does not fit space in log.recent_closed, bypassing
72 # required wait for space there (thanks to MTR_LOG_NO_REDO).
73 #
74 #
75 # 5. Resume the old mtr which blocked the log.recent_closed.tail() from advancing.
76 #
77 #
78 # 6. Disconnect pending connections.
79 #
80 #
81 # 7. Perform standard mtrs, just to ensure all is fine.
82 #
83 include/assert.inc [We must not have oldest_lsn_approx < checkpoint_lsn.]
84 # Kill and restart
85 #
86 # Scenario 1.2.1 - No-redo dirtied page inserted too quickly.
87 # Empty flush list when it is inserted.
88 # The page is no-redo dirtied first time.
89 #
90 #
91 # 0. Block log.recent_closed.tail() from advancing by stopping single mtr
92 # just before it inserts its dirty pages to flush list.
93 #
94 #
95 # 1. Frozen recent_closed.
96 #
97 #
98 # 2. Optionally modify no-redo dirtied page (in which case next time we remodify it)
99 # Note it will be flushed meanwhile anyway.
100 #
101 #
102 # 3. Start transactions and open tables.
103 #
104 #
105 # 4. Keep modifying pages, stop just before we reach limit in recent_closed.
106 #
107 #
108 # 5. Flush all dirty pages to make flush list empty.
109 #
110 #
111 # 6. Execute multiple mtr, stopping them before they add any pages...
112 #
113 #
114 # 7. Wait until we run out of free space in recent_closed...
115 #
116 #
117 # 8. Execute mini-transactions with no-redo dirtied pages (MTR_LOG_NO_REDO).
118 # When bug was present, such mtr inserted pages with too big lsn using
119 # current lsn, which does not fit space in log.recent_closed, bypassing
120 # required wait for space there (thanks to MTR_LOG_NO_REDO).
121 #
122 #
123 # 9. Resume log.recent_closed
124 #
125 #
126 # 10. Finish pending mini-transactions
127 #
128 #
129 # 11. Perform standard mtrs, just to ensure all is fine.
130 #
131 include/assert.inc [We must not have oldest_lsn_approx < checkpoint_lsn.]
132 # Kill and restart
133 #
134 # Scenario 1.2.2 - No-redo dirtied page inserted too quickly.
135 # Empty flush list when it is inserted.
136 # The page is no-redo dirtied second time,
137 # but it was flushed (empty flush list).
138 #
139 #
140 # 0. Block log.recent_closed.tail() from advancing by stopping single mtr
141 # just before it inserts its dirty pages to flush list.
142 #
143 #
144 # 1. Frozen recent_closed.
145 #
146 #
147 # 2. Optionally modify no-redo dirtied page (in which case next time we remodify it)
148 # Note it will be flushed meanwhile anyway.
149 #
150 #
151 # 3. Start transactions and open tables.
152 #
153 #
154 # 4. Keep modifying pages, stop just before we reach limit in recent_closed.
155 #
156 #
157 # 5. Flush all dirty pages to make flush list empty.
158 #
159 #
160 # 6. Execute multiple mtr, stopping them before they add any pages...
161 #
162 #
163 # 7. Wait until we run out of free space in recent_closed...
164 #
165 #
166 # 8. Execute mini-transactions with no-redo dirtied pages (MTR_LOG_NO_REDO).
167 # When bug was present, such mtr inserted pages with too big lsn using
168 # current lsn, which does not fit space in log.recent_closed, bypassing
169 # required wait for space there (thanks to MTR_LOG_NO_REDO).
170 #
171 #
172 # 9. Resume log.recent_closed
173 #
174 #
175 # 10. Finish pending mini-transactions
176 #
177 #
178 # 11. Perform standard mtrs, just to ensure all is fine.
179 #
180 include/assert.inc [We must not have oldest_lsn_approx < checkpoint_lsn.]
181 # Kill and restart
182 #
183 # Scenario 2.1.1 - No-redo dirtied page inserted too late.
184 # Non-empty flush list when it is inserted.
185 # The page is no-redo dirtied first time.
186 #
187 #
188 # 0. Freeze single no-redo mtr (MTR_LOG_NO_REDO) in mtr0mtr.cc, just
189 # before it inserts the dirty page to flush list.
190 #
191 #
192 # 1. Allow user threads to execute many redo-based mini-transactions,
193 # until current lsn is greater than init_lsn by more than capacity
194 # of log.recent_closed.
195 #
196 #
197 # 2. Optionally flush all dirty pages.
198 #
199 #
200 # 3. Resume the frozen no-redo mtr which seems to be too old.
201 #
202 #
203 # 4. Finish pending insertions in redo-based mini-transactions.
204 #
205 #
206 # 5. Execute two standard mini-transactions to ensure all is fine.
207 #
208 include/assert.inc [We must not have oldest_lsn_approx < checkpoint_lsn.]
209 # Kill and restart
210 #
211 # Scenario 2.1.2 - No-redo dirtied page inserted too late.
212 # Non-empty flush list when it is inserted.
213 # The page is no-redo dirtied second time.
214 #
215 #
216 # 0. Freeze single no-redo mtr (MTR_LOG_NO_REDO) in mtr0mtr.cc, just
217 # before it inserts the dirty page to flush list.
218 #
219 #
220 # 1. Allow user threads to execute many redo-based mini-transactions,
221 # until current lsn is greater than init_lsn by more than capacity
222 # of log.recent_closed.
223 #
224 #
225 # 2. Optionally flush all dirty pages.
226 #
227 #
228 # 3. Resume the frozen no-redo mtr which seems to be too old.
229 #
230 #
231 # 4. Finish pending insertions in redo-based mini-transactions.
232 #
233 #
234 # 5. Execute two standard mini-transactions to ensure all is fine.
235 #
236 include/assert.inc [We must not have oldest_lsn_approx < checkpoint_lsn.]
237 # Kill and restart
238 #
239 # Scenario 2.2.1 - No-redo dirtied page inserted too late.
240 # Empty flush list when it is inserted.
241 # The page is no-redo dirtied first time.
242 #
243 #
244 # 0. Freeze single no-redo mtr (MTR_LOG_NO_REDO) in mtr0mtr.cc, just
245 # before it inserts the dirty page to flush list.
246 #
247 #
248 # 1. Allow user threads to execute many redo-based mini-transactions,
249 # until current lsn is greater than init_lsn by more than capacity
250 # of log.recent_closed.
251 #
252 #
253 # 2. Optionally flush all dirty pages.
254 #
255 Flush all dirty pages...
256 #
257 # 3. Resume the frozen no-redo mtr which seems to be too old.
258 #
259 #
260 # 4. Finish pending insertions in redo-based mini-transactions.
261 #
262 #
263 # 5. Execute two standard mini-transactions to ensure all is fine.
264 #
265 include/assert.inc [We must not have oldest_lsn_approx < checkpoint_lsn.]
266 # Kill and restart
267 #
268 # Scenario 2.2.2 - No-redo dirtied page inserted too late.
269 # Empty flush list when it is inserted.
270 # The page is no-redo dirtied second time,
271 # but it was flushed (empty flush list).
272 #
273 #
274 # 0. Freeze single no-redo mtr (MTR_LOG_NO_REDO) in mtr0mtr.cc, just
275 # before it inserts the dirty page to flush list.
276 #
277 #
278 # 1. Allow user threads to execute many redo-based mini-transactions,
279 # until current lsn is greater than init_lsn by more than capacity
280 # of log.recent_closed.
281 #
282 #
283 # 2. Optionally flush all dirty pages.
284 #
285 Flush all dirty pages...
286 #
287 # 3. Resume the frozen no-redo mtr which seems to be too old.
288 #
289 #
290 # 4. Finish pending insertions in redo-based mini-transactions.
291 #
292 #
293 # 5. Execute two standard mini-transactions to ensure all is fine.
294 #
295 include/assert.inc [We must not have oldest_lsn_approx < checkpoint_lsn.]
296 # Kill and restart
297 #
298 # Scenario 3 - Execute in parallel many redo and no-redo mini-transactions.
299 # Freeze and unfreeze recent_closed every 1s.
300 #
301 #
302 # 0. Freeze log.recent_closed.
303 #
304 #
305 # 1. Start threads doing redo-based mtr and threads doing noredo-based mtr.
306 #
307 #
308 # 2. Freeze/unfreeze log_recent_closed every 1s
309 #
310 Iterations left: 10 - unfreeze...
311 Iterations left: 10 - freeze...
312 Iterations left: 9 - unfreeze...
313 Iterations left: 9 - freeze...
314 Iterations left: 8 - unfreeze...
315 Iterations left: 8 - freeze...
316 Iterations left: 7 - unfreeze...
317 Iterations left: 7 - freeze...
318 Iterations left: 6 - unfreeze...
319 Iterations left: 6 - freeze...
320 Iterations left: 5 - unfreeze...
321 Iterations left: 5 - freeze...
322 Iterations left: 4 - unfreeze...
323 Iterations left: 4 - freeze...
324 Iterations left: 3 - unfreeze...
325 Iterations left: 3 - freeze...
326 Iterations left: 2 - unfreeze...
327 Iterations left: 2 - freeze...
328 Iterations left: 1 - unfreeze...
329 Iterations left: 1 - freeze...
330 Finished all iterations - unfreeze...
331 #
332 # 3. Finish pending insertions.
333 #
334 include/assert.inc [We must not have oldest_lsn_approx < checkpoint_lsn.]
335 # Kill and restart
336 #
337 # CLEANUP
338 #