| 1 |
# include/socket_summary_check.inc |
| 2 |
# |
| 3 |
# Auxiliary routine to be sourced by socket_summary_by_instance_func.test |
| 4 |
# or other routines sourced within this script. |
| 5 |
# |
| 6 |
# Purpose |
| 7 |
# Various checks for the content of the table socket_summary_by_instance. |
| 8 |
# |
| 9 |
# It is intentional that we do not try to cram as much checks as possible into |
| 10 |
# one single SQL statement. |
| 11 |
# Reasons: |
| 12 |
# - We check performance_schema here and NOT something like optimizer. |
| 13 |
# - This test should work even if some other feature has become buggy. |
| 14 |
# - In case some check gives unexpected results than we print the |
| 15 |
# relevant content of the table and the values which we expect. |
| 16 |
# In case of all checks in one statement such a printout would be too huge. |
| 17 |
# |
| 18 |
# IMPORTANT: |
| 19 |
# The maximum number of rows which the table socket_summary_by_instance |
| 20 |
# can keep is limited via the system variables max_socket_classes and |
| 21 |
# max_socket_instances. We are running with the default values here. |
| 22 |
# They are sufficient high so that these limits cannot harm the current test. |
| 23 |
# FIXME: Check at the beginning of the test that the limits are sufficient |
| 24 |
# for the current test. |
| 25 |
# |
| 26 |
|
| 27 |
--disable_query_log |
| 28 |
# Insert the current state into mysqltest.my_socket_summary_by_instance. |
| 29 |
eval $insert_after; |
| 30 |
--enable_query_log |
| 31 |
--enable_result_log |
| 32 |
|
| 33 |
|
| 34 |
# 1. The content of socket_summary_by_instance must be consistent to the |
| 35 |
# content of socket_instances |
| 36 |
#======================================================================= |
| 37 |
let $part1= |
| 38 |
FROM performance_schema.socket_summary_by_instance |
| 39 |
WHERE (EVENT_NAME,OBJECT_INSTANCE_BEGIN) |
| 40 |
NOT IN (SELECT EVENT_NAME,OBJECT_INSTANCE_BEGIN |
| 41 |
FROM performance_schema.socket_instances); |
| 42 |
if(`SELECT COUNT(*) $part1`) |
| 43 |
{ |
| 44 |
--echo # There is an inconsistency between the content of the tables |
| 45 |
--echo # socket_instances and socket_summary_by_instance |
| 46 |
--echo # |
| 47 |
eval |
| 48 |
SELECT 'not in socket_instances' AS state, EVENT_NAME, OBJECT_INSTANCE_BEGIN |
| 49 |
$part1; |
| 50 |
} |
| 51 |
|
| 52 |
--vertical_results |
| 53 |
|
| 54 |
|
| 55 |
# 2. The computation of statistics must be roughly correct. |
| 56 |
# |
| 57 |
# If we run this check sufficient frequent than AVG_TIMER_* can be removed from other checks. |
| 58 |
#=============================================================================================== |
| 59 |
let $my_lo= 0.98; |
| 60 |
let $my_hi= 1.02; |
| 61 |
|
| 62 |
let $my_rules= |
| 63 |
COUNT_STAR * AVG_TIMER_WAIT BETWEEN SUM_TIMER_WAIT * $my_lo AND SUM_TIMER_WAIT * $my_hi AND |
| 64 |
COUNT_READ * AVG_TIMER_READ BETWEEN SUM_TIMER_READ * $my_lo AND SUM_TIMER_READ * $my_hi AND |
| 65 |
COUNT_WRITE * AVG_TIMER_WRITE BETWEEN SUM_TIMER_WRITE * $my_lo AND SUM_TIMER_WRITE * $my_hi AND |
| 66 |
COUNT_MISC * AVG_TIMER_MISC BETWEEN SUM_TIMER_MISC * $my_lo AND SUM_TIMER_MISC * $my_hi; |
| 67 |
|
| 68 |
let $part= |
| 69 |
SUM_TIMER_WAIT * $my_lo, COUNT_STAR * AVG_TIMER_WAIT, SUM_TIMER_WAIT * $my_hi, |
| 70 |
COUNT_STAR, SUM_TIMER_WAIT, AVG_TIMER_WAIT, |
| 71 |
SUM_TIMER_READ * $my_lo, COUNT_READ * AVG_TIMER_READ, SUM_TIMER_READ * $my_hi, |
| 72 |
COUNT_READ, SUM_TIMER_READ, AVG_TIMER_READ, |
| 73 |
SUM_TIMER_WRITE * $my_lo, COUNT_WRITE * AVG_TIMER_WRITE, SUM_TIMER_WRITE * $my_hi, |
| 74 |
COUNT_WRITE, SUM_TIMER_WRITE, AVG_TIMER_WRITE, |
| 75 |
SUM_TIMER_MISC * $my_lo, COUNT_MISC * AVG_TIMER_MISC, SUM_TIMER_MISC * $my_hi, |
| 76 |
COUNT_MISC, SUM_TIMER_MISC, AVG_TIMER_MISC; |
| 77 |
|
| 78 |
if(`SELECT SUM($my_rules) <> COUNT(*) FROM mysqltest.my_socket_summary_by_instance |
| 79 |
WHERE pk = 'After'`) |
| 80 |
{ |
| 81 |
--echo # The statistics looks suspicious. |
| 82 |
--echo # We expect |
| 83 |
--echo # $my_rules |
| 84 |
eval |
| 85 |
SELECT EVENT_NAME, OBJECT_INSTANCE_BEGIN, |
| 86 |
$part |
| 87 |
FROM mysqltest.my_socket_summary_by_instance |
| 88 |
WHERE pk = 'After' AND NOT ($my_rules) |
| 89 |
ORDER BY EVENT_NAME, OBJECT_INSTANCE_BEGIN; |
| 90 |
let $print_details= 1; |
| 91 |
} |
| 92 |
|
| 93 |
|
| 94 |
# 3. Check the relation between AVG_*, MIN_TIMER_* and MAX_TIMER_* |
| 95 |
# |
| 96 |
# If we run this check sufficient frequent than only the following |
| 97 |
# additional checks are required: |
| 98 |
# a) If (SUM_TIMER_*(new) - SUM_TIMER_*(old) < MIN_TIMER_*(old)) |
| 99 |
# than MIN_TIMER_*(new) = SUM_TIMER_*(new) - SUM_TIMER_*(old). |
| 100 |
# b) If (SUM_TIMER_*(new) - SUM_TIMER_*(old) > MAX_TIMER_*(old)) |
| 101 |
# than MAX_TIMER_*(new) = SUM_TIMER_*(new) - SUM_TIMER_*(old). |
| 102 |
# in order to remove MIN_TIMER_* and MAX_TIMER_* from other checks |
| 103 |
# Between the states "new" and "old" must be exact one statement. |
| 104 |
#----------------------------------------------------------------------------------------------- |
| 105 |
let $my_rules= |
| 106 |
AVG_TIMER_WAIT >= MIN_TIMER_WAIT AND MAX_TIMER_WAIT >= AVG_TIMER_WAIT AND |
| 107 |
AVG_TIMER_READ >= MIN_TIMER_READ AND MAX_TIMER_READ >= AVG_TIMER_READ AND |
| 108 |
AVG_TIMER_WRITE >= MIN_TIMER_WRITE AND MAX_TIMER_WRITE >= AVG_TIMER_WRITE AND |
| 109 |
AVG_TIMER_MISC >= MIN_TIMER_MISC AND MAX_TIMER_MISC >= AVG_TIMER_MISC; |
| 110 |
|
| 111 |
let $part= |
| 112 |
MIN_TIMER_WAIT, AVG_TIMER_WAIT, MAX_TIMER_WAIT, |
| 113 |
MIN_TIMER_READ, AVG_TIMER_READ, MAX_TIMER_READ, |
| 114 |
MIN_TIMER_WRITE, AVG_TIMER_WRITE, MAX_TIMER_WRITE, |
| 115 |
MIN_TIMER_MISC, AVG_TIMER_MISC, MAX_TIMER_MISC; |
| 116 |
|
| 117 |
if(`SELECT SUM($my_rules) <> COUNT(*) FROM mysqltest.my_socket_summary_by_instance |
| 118 |
WHERE pk = 'After'`) |
| 119 |
{ |
| 120 |
--echo # The statistics looks suspicious. |
| 121 |
--echo # We expect |
| 122 |
--echo # $my_rules |
| 123 |
eval |
| 124 |
SELECT EVENT_NAME, OBJECT_INSTANCE_BEGIN, |
| 125 |
$part |
| 126 |
FROM mysqltest.my_socket_summary_by_instance |
| 127 |
WHERE pk = 'After' AND NOT ($my_rules) |
| 128 |
ORDER BY EVENT_NAME, OBJECT_INSTANCE_BEGIN; |
| 129 |
let $print_details= 1; |
| 130 |
} |
| 131 |
|
| 132 |
|
| 133 |
# 4. Check the aggregate columns COUNT_STAR and SUM_TIMER_WAIT |
| 134 |
# |
| 135 |
# The specification says: |
| 136 |
# The columns COUNT_STAR, SUM/MIN/AVG/MAX TIMER_WAIT aggregates all operations. |
| 137 |
# |
| 138 |
# If we run this check sufficient frequent than COUNT_STAR and SUM_TIMER_WAIT |
| 139 |
# can be removed from other checks. |
| 140 |
#--------------------------------------------------------------------------------- |
| 141 |
let $my_rules= |
| 142 |
COUNT_STAR = COUNT_READ + COUNT_WRITE + COUNT_MISC AND |
| 143 |
SUM_TIMER_WAIT = SUM_TIMER_READ + SUM_TIMER_WRITE + SUM_TIMER_MISC; |
| 144 |
|
| 145 |
let $part= |
| 146 |
COUNT_STAR, COUNT_READ + COUNT_WRITE + COUNT_MISC, COUNT_READ, COUNT_WRITE, COUNT_MISC, |
| 147 |
SUM_TIMER_WAIT, SUM_TIMER_READ + SUM_TIMER_WRITE + SUM_TIMER_MISC, SUM_TIMER_READ, |
| 148 |
SUM_TIMER_WRITE, SUM_TIMER_MISC; |
| 149 |
|
| 150 |
if(`SELECT SUM($my_rules) <> COUNT(*) FROM mysqltest.my_socket_summary_by_instance |
| 151 |
WHERE pk = 'After'`) |
| 152 |
{ |
| 153 |
--echo # The statistics looks suspicious. |
| 154 |
--echo # We expect |
| 155 |
--echo # $my_rules |
| 156 |
--echo # |
| 157 |
eval |
| 158 |
SELECT EVENT_NAME, OBJECT_INSTANCE_BEGIN, |
| 159 |
$part |
| 160 |
FROM mysqltest.my_socket_summary_by_instance |
| 161 |
WHERE pk = 'After' |
| 162 |
ORDER BY EVENT_NAME, OBJECT_INSTANCE_BEGIN; |
| 163 |
let $print_details= 1; |
| 164 |
} |
| 165 |
|
| 166 |
|
| 167 |
# 5. Check the aggregate column MIN_TIMER_WAIT |
| 168 |
# |
| 169 |
# If we run this check sufficient frequent than MIN_TIMER_WAIT |
| 170 |
# can be removed from other checks. |
| 171 |
#--------------------------------------------------------------------------------- |
| 172 |
let $my_rules= |
| 173 |
MIN_TIMER_WAIT >= mysqltest.min_of_triple(MIN_TIMER_READ,MIN_TIMER_WRITE,MIN_TIMER_MISC); |
| 174 |
|
| 175 |
let $part= |
| 176 |
MIN_TIMER_WAIT, |
| 177 |
mysqltest.min_of_triple(MIN_TIMER_READ, MIN_TIMER_WRITE, MIN_TIMER_MISC) AS "Min_of_Triple", |
| 178 |
MIN_TIMER_READ, MIN_TIMER_WRITE, MIN_TIMER_MISC; |
| 179 |
|
| 180 |
if(`SELECT SUM($my_rules) <> COUNT(*) FROM mysqltest.my_socket_summary_by_instance |
| 181 |
WHERE pk = 'After'`) |
| 182 |
{ |
| 183 |
--echo # The statistics looks suspicious. |
| 184 |
--echo # We expect |
| 185 |
--echo # $my_rules |
| 186 |
--echo # |
| 187 |
eval |
| 188 |
SELECT EVENT_NAME, OBJECT_INSTANCE_BEGIN, |
| 189 |
$part |
| 190 |
FROM mysqltest.my_socket_summary_by_instance |
| 191 |
WHERE pk = 'After' |
| 192 |
ORDER BY EVENT_NAME, OBJECT_INSTANCE_BEGIN; |
| 193 |
let $print_details= 1; |
| 194 |
} |
| 195 |
|
| 196 |
|
| 197 |
# 6. Check the aggregate column MAX_TIMER_WAIT |
| 198 |
# |
| 199 |
# If we run this check sufficient frequent than MAX_TIMER_WAIT |
| 200 |
# can be removed from other checks. |
| 201 |
#--------------------------------------------------------------------------------- |
| 202 |
let $my_rules= |
| 203 |
MAX_TIMER_WAIT >= mysqltest.max_of_triple(MAX_TIMER_READ,MAX_TIMER_WRITE,MAX_TIMER_MISC); |
| 204 |
|
| 205 |
let $part= |
| 206 |
MAX_TIMER_WAIT, |
| 207 |
mysqltest.max_of_triple(MAX_TIMER_READ, MAX_TIMER_WRITE, MAX_TIMER_MISC) AS "Max_of_Triple", |
| 208 |
MAX_TIMER_READ, MAX_TIMER_WRITE, MAX_TIMER_MISC; |
| 209 |
|
| 210 |
if(`SELECT SUM($my_rules) <> COUNT(*) FROM mysqltest.my_socket_summary_by_instance |
| 211 |
WHERE pk = 'After'`) |
| 212 |
{ |
| 213 |
--echo # The statistics looks suspicious. |
| 214 |
--echo # We expect |
| 215 |
--echo # $my_rules |
| 216 |
--echo # |
| 217 |
eval |
| 218 |
SELECT EVENT_NAME, OBJECT_INSTANCE_BEGIN, |
| 219 |
$part |
| 220 |
FROM mysqltest.my_socket_summary_by_instance |
| 221 |
WHERE pk = 'After' |
| 222 |
ORDER BY EVENT_NAME, OBJECT_INSTANCE_BEGIN; |
| 223 |
let $print_details= 1; |
| 224 |
} |
| 225 |
|
| 226 |
--horizontal_results |
| 227 |
|