p_s/file_summary_by_instance/file_summary_by_instance_row.go

   1 // This file contains the library routines for managing the
   2 // file_summary_by_instance table.
   3 package file_summary_by_instance
   4
   5 import (
   6         "database/sql"
   7         "fmt"
   8         "log"
   9         "regexp"
  10         "sort"
  11         "time"
  12
  13         "github.com/sjmudd/pstop/key_value_cache"
  14         "github.com/sjmudd/pstop/lib"
  15 )
  16
  17 /*
  18 CREATE TABLE `file_summary_by_instance` (
  19   `FILE_NAME` varchar(512) NOT NULL,
  20   `EVENT_NAME` varchar(128) NOT NULL,                           // not collected
  21   `OBJECT_INSTANCE_BEGIN` bigint(20) unsigned NOT NULL,         // not collected
  22   `COUNT_STAR` bigint(20) unsigned NOT NULL,
  23   `SUM_TIMER_WAIT` bigint(20) unsigned NOT NULL,
  24   `MIN_TIMER_WAIT` bigint(20) unsigned NOT NULL,
  25   `AVG_TIMER_WAIT` bigint(20) unsigned NOT NULL,
  26   `MAX_TIMER_WAIT` bigint(20) unsigned NOT NULL,
  27   `COUNT_READ` bigint(20) unsigned NOT NULL,
  28   `SUM_TIMER_READ` bigint(20) unsigned NOT NULL,
  29   `MIN_TIMER_READ` bigint(20) unsigned NOT NULL,
  30   `AVG_TIMER_READ` bigint(20) unsigned NOT NULL,
  31   `MAX_TIMER_READ` bigint(20) unsigned NOT NULL,
  32   `SUM_NUMBER_OF_BYTES_READ` bigint(20) NOT NULL,
  33   `COUNT_WRITE` bigint(20) unsigned NOT NULL,
  34   `SUM_TIMER_WRITE` bigint(20) unsigned NOT NULL,
  35   `MIN_TIMER_WRITE` bigint(20) unsigned NOT NULL,
  36   `AVG_TIMER_WRITE` bigint(20) unsigned NOT NULL,
  37   `MAX_TIMER_WRITE` bigint(20) unsigned NOT NULL,
  38   `SUM_NUMBER_OF_BYTES_WRITE` bigint(20) NOT NULL,
  39   `COUNT_MISC` bigint(20) unsigned NOT NULL,
  40   `SUM_TIMER_MISC` bigint(20) unsigned NOT NULL,
  41   `MIN_TIMER_MISC` bigint(20) unsigned NOT NULL,
  42   `AVG_TIMER_MISC` bigint(20) unsigned NOT NULL,
  43   `MAX_TIMER_MISC` bigint(20) unsigned NOT NULL
  44 ) ENGINE=PERFORMANCE_SCHEMA DEFAULT CHARSET=utf8
  45 1 row in set (0.00 sec)
  46 */
  47
  48 //     foo/../bar --> foo/bar   perl: $new =~ s{[^/]+/\.\./}{/};
  49 //     /./        --> /         perl: $new =~ s{/\./}{};
  50 //     //         --> /         perl: $new =~ s{//}{/};
  51 const (
  52         re_encoded = `@(\d{4})` // FIXME - add me to catch @0024 --> $ for example
  53 )
  54
  55 var (
  56         re_one_or_the_other    *regexp.Regexp = regexp.MustCompile(`/(\.)?/`)
  57         re_slash_dot_dot_slash *regexp.Regexp = regexp.MustCompile(`[^/]+/\.\./`)
  58         re_table_file          *regexp.Regexp = regexp.MustCompile(`/([^/]+)/([^/]+)\.(frm|ibd|MYD|MYI|CSM|CSV|par)$`)
  59         re_temp_table          *regexp.Regexp = regexp.MustCompile(`#sql-[0-9_]+`)
  60         re_part_table          *regexp.Regexp = regexp.MustCompile(`(.+)#P#p(\d+|MAX)`)
  61         re_ibdata              *regexp.Regexp = regexp.MustCompile(`/ibdata\d+$`)
  62         re_redo_log            *regexp.Regexp = regexp.MustCompile(`/ib_logfile\d+$`)
  63         re_binlog              *regexp.Regexp = regexp.MustCompile(`/binlog\.(\d{6}|index)$`)
  64         re_db_opt              *regexp.Regexp = regexp.MustCompile(`/db\.opt$`)
  65         re_slowlog             *regexp.Regexp = regexp.MustCompile(`/slowlog$`)
  66         re_auto_cnf            *regexp.Regexp = regexp.MustCompile(`/auto\.cnf$`)
  67         re_pid_file            *regexp.Regexp = regexp.MustCompile(`/[^/]+\.pid$`)
  68         re_error_msg           *regexp.Regexp = regexp.MustCompile(`/share/[^/]+/errmsg\.sys$`)
  69         re_charset             *regexp.Regexp = regexp.MustCompile(`/share/charsets/Index\.xml$`)
  70         re_dollar              *regexp.Regexp = regexp.MustCompile(`@0024`) // FIXME - add me to catch @0024 --> $ (specific case)
  71
  72         cache key_value_cache.KeyValueCache
  73 )
  74
  75 type file_summary_by_instance_row struct {
  76         FILE_NAME string
  77
  78         COUNT_STAR  uint64
  79         COUNT_READ  uint64
  80         COUNT_WRITE uint64
  81         COUNT_MISC  uint64
  82
  83         SUM_TIMER_WAIT  uint64
  84         SUM_TIMER_READ  uint64
  85         SUM_TIMER_WRITE uint64
  86         SUM_TIMER_MISC  uint64
  87
  88         SUM_NUMBER_OF_BYTES_READ  uint64
  89         SUM_NUMBER_OF_BYTES_WRITE uint64
  90 }
  91
  92 // represents a table or set of rows
  93 type file_summary_by_instance_rows []file_summary_by_instance_row
  94
  95 // Return the name using the FILE_NAME attribute.
  96 func (r *file_summary_by_instance_row) name() string {
  97         return r.FILE_NAME
  98 }
  99
 100 // Return a formatted pretty name for the row.
 101 func (r *file_summary_by_instance_row) pretty_name() string {
 102         s := r.name()
 103         if len(s) > 30 {
 104                 s = s[:29]
 105         }
 106         return s
 107 }
 108
 109 func (r *file_summary_by_instance_row) headings() string {
 110         return fmt.Sprintf("%-30s %10s %6s|%6s %6s %6s|%8s %8s|%8s %6s %6s %6s",
 111                 "Table Name",
 112                 "Latency",
 113                 "%",
 114                 "Read",
 115                 "Write",
 116                 "Misc",
 117                 "Rd bytes",
 118                 "Wr bytes",
 119                 "Ops",
 120                 "R Ops",
 121                 "W Ops",
 122                 "M Ops")
 123 }
 124
 125 // generate a printable result
 126 func (row *file_summary_by_instance_row) row_content(totals file_summary_by_instance_row) string {
 127         var name string = row.pretty_name()
 128
 129         // We assume that if COUNT_STAR = 0 then there's no data at all...
 130         // when we have no data we really don't want to show the name either.
 131         if row.COUNT_STAR == 0 && name != "Totals" {
 132                 name = ""
 133         }
 134
 135         return fmt.Sprintf("%-30s %10s %6s|%6s %6s %6s|%8s %8s|%8s %6s %6s %6s",
 136                 name,
 137                 lib.FormatTime(row.SUM_TIMER_WAIT),
 138                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_WAIT, totals.SUM_TIMER_WAIT)),
 139                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_READ, row.SUM_TIMER_WAIT)),
 140                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_WRITE, row.SUM_TIMER_WAIT)),
 141                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_MISC, row.SUM_TIMER_WAIT)),
 142                 lib.FormatAmount(row.SUM_NUMBER_OF_BYTES_READ),
 143                 lib.FormatAmount(row.SUM_NUMBER_OF_BYTES_WRITE),
 144                 lib.FormatAmount(row.COUNT_STAR),
 145                 lib.FormatPct(lib.MyDivide(row.COUNT_READ, row.COUNT_STAR)),
 146                 lib.FormatPct(lib.MyDivide(row.COUNT_WRITE, row.COUNT_STAR)),
 147                 lib.FormatPct(lib.MyDivide(row.COUNT_MISC, row.COUNT_STAR)))
 148 }
 149
 150 func (this *file_summary_by_instance_row) add(other file_summary_by_instance_row) {
 151         this.COUNT_STAR += other.COUNT_STAR
 152         this.COUNT_READ += other.COUNT_READ
 153         this.COUNT_WRITE += other.COUNT_WRITE
 154         this.COUNT_MISC += other.COUNT_MISC
 155
 156         this.SUM_TIMER_WAIT += other.SUM_TIMER_WAIT
 157         this.SUM_TIMER_READ += other.SUM_TIMER_READ
 158         this.SUM_TIMER_WRITE += other.SUM_TIMER_WRITE
 159         this.SUM_TIMER_MISC += other.SUM_TIMER_MISC
 160
 161         this.SUM_NUMBER_OF_BYTES_READ += other.SUM_NUMBER_OF_BYTES_READ
 162         this.SUM_NUMBER_OF_BYTES_WRITE += other.SUM_NUMBER_OF_BYTES_WRITE
 163 }
 164
 165 func (this *file_summary_by_instance_row) subtract(other file_summary_by_instance_row) {
 166         this.COUNT_STAR -= other.COUNT_STAR
 167         this.COUNT_READ -= other.COUNT_READ
 168         this.COUNT_WRITE -= other.COUNT_WRITE
 169         this.COUNT_MISC -= other.COUNT_MISC
 170
 171         this.SUM_TIMER_WAIT -= other.SUM_TIMER_WAIT
 172         this.SUM_TIMER_READ -= other.SUM_TIMER_READ
 173         this.SUM_TIMER_WRITE -= other.SUM_TIMER_WRITE
 174         this.SUM_TIMER_MISC -= other.SUM_TIMER_MISC
 175
 176         this.SUM_NUMBER_OF_BYTES_READ -= other.SUM_NUMBER_OF_BYTES_READ
 177         this.SUM_NUMBER_OF_BYTES_WRITE -= other.SUM_NUMBER_OF_BYTES_WRITE
 178 }
 179
 180 // return the totals of a slice of rows
 181 func (t file_summary_by_instance_rows) totals() file_summary_by_instance_row {
 182         var totals file_summary_by_instance_row
 183         totals.FILE_NAME = "Totals"
 184
 185         for i := range t {
 186                 totals.add(t[i])
 187         }
 188
 189         return totals
 190 }
 191
 192 // clean up the given path reducing redundant stuff and return the clean path
 193 func cleanup_path(path string) string {
 194
 195         for {
 196                 orig_path := path
 197                 path = re_one_or_the_other.ReplaceAllString(path, "/")
 198                 path = re_slash_dot_dot_slash.ReplaceAllString(path, "/")
 199                 if orig_path == path { // no change so give up
 200                         break
 201                 }
 202         }
 203
 204         return path
 205 }
 206
 207 // From the original FILE_NAME we want to generate a simpler name to use.
 208 // This simpler name may also merge several different filenames into one.
 209 func (t file_summary_by_instance_row) simple_name(global_variables map[string]string) string {
 210
 211         path := t.FILE_NAME
 212
 213         if cached_result, err := cache.Get(path); err == nil {
 214                 return cached_result
 215         }
 216
 217         // @0024 --> $ (should do this more generically)
 218         path = re_dollar.ReplaceAllLiteralString(path, "$")
 219
 220         // this should probably be ordered from most expected regexp to least
 221         if m1 := re_table_file.FindStringSubmatch(path); m1 != nil {
 222                 // we may match temporary tables so check for them
 223                 if m2 := re_temp_table.FindStringSubmatch(m1[2]); m2 != nil {
 224                         return cache.Put(path, "<temp_table>")
 225                 }
 226
 227                 // we may match partitioned tables so check for them
 228                 if m3 := re_part_table.FindStringSubmatch(m1[2]); m3 != nil {
 229                         return cache.Put(path, m1[1]+"."+m3[1]) // <schema>.<table> (less partition info)
 230                 }
 231
 232                 return cache.Put(path, m1[1]+"."+m1[2]) // <schema>.<table>
 233         }
 234         if re_ibdata.MatchString(path) == true {
 235                 return cache.Put(path, "<ibdata>")
 236         }
 237         if re_redo_log.MatchString(path) == true {
 238                 return cache.Put(path, "<redo_log>")
 239         }
 240         if re_binlog.MatchString(path) == true {
 241                 return cache.Put(path, "<binlog>")
 242         }
 243         if re_db_opt.MatchString(path) == true {
 244                 return cache.Put(path, "<db_opt>")
 245         }
 246         if re_slowlog.MatchString(path) == true {
 247                 return cache.Put(path, "<slow_log>")
 248         }
 249         if re_auto_cnf.MatchString(path) == true {
 250                 return cache.Put(path, "<auto_cnf>")
 251         }
 252         // relay logs are a bit complicated. If a full path then easy to
 253         // identify,but if a relative path we may need to add $datadir,
 254         // but also if as I do we have a ../blah/somewhere/path then we
 255         // need to make it match too.
 256         if len(global_variables["relay_log"]) > 0 {
 257                 relay_log := global_variables["relay_log"]
 258                 if relay_log[0] != '/' { // relative path
 259                         relay_log = cleanup_path(global_variables["datadir"] + relay_log) // datadir always ends in /
 260                 }
 261                 re_relay_log := relay_log + `\.(\d{6}|index)$`
 262                 if regexp.MustCompile(re_relay_log).MatchString(path) == true {
 263                         return cache.Put(path, "<relay_log>")
 264                 }
 265         }
 266         if re_pid_file.MatchString(path) == true {
 267                 return cache.Put(path, "<pid_file>")
 268         }
 269         if re_error_msg.MatchString(path) == true {
 270                 return cache.Put(path, "<errmsg>")
 271         }
 272         if re_charset.MatchString(path) == true {
 273                 return cache.Put(path, "<charset>")
 274         }
 275         // clean up datadir to <datadir>
 276         if len(global_variables["datadir"]) > 0 {
 277                 re_datadir := regexp.MustCompile("^" + global_variables["datadir"])
 278                 path = re_datadir.ReplaceAllLiteralString(path, "<datadir>/")
 279         }
 280
 281         return cache.Put(path, path)
 282 }
 283
 284 // Convert the imported "table" to a merged one with merged data.
 285 // Combine all entries with the same "FILE_NAME" by adding their values.
 286 func merge_by_table_name(orig file_summary_by_instance_rows, global_variables map[string]string) file_summary_by_instance_rows {
 287         start := time.Now()
 288         t := make(file_summary_by_instance_rows, 0, len(orig))
 289
 290         m := make(map[string]file_summary_by_instance_row)
 291
 292         // iterate over source table
 293         for i := range orig {
 294                 var file_name string
 295                 var new_row file_summary_by_instance_row
 296                 orig_row := orig[i]
 297
 298                 if orig_row.COUNT_STAR > 0 {
 299                         file_name = orig_row.simple_name(global_variables)
 300
 301                         // check if we have an entry in the map
 302                         if _, found := m[file_name]; found {
 303                                 new_row = m[file_name]
 304                         } else {
 305                                 new_row.FILE_NAME = file_name
 306                         }
 307                         new_row.add(orig_row)
 308                         m[file_name] = new_row // update the map with the new value
 309                 }
 310         }
 311
 312         // add the map contents back into the table
 313         for _, row := range m {
 314                 t = append(t, row)
 315         }
 316
 317         lib.Logger.Println("merge_by_table_name() took:", time.Duration(time.Since(start)).String())
 318         return t
 319 }
 320
 321 // Select the raw data from the database into file_summary_by_instance_rows
 322 // - filter out empty values
 323 // - merge rows with the same name into a single row
 324 // - change FILE_NAME into a more descriptive value.
 325 func select_fsbi_rows(dbh *sql.DB) file_summary_by_instance_rows {
 326         var t file_summary_by_instance_rows
 327         start := time.Now()
 328
 329         sql := "SELECT FILE_NAME, COUNT_STAR, SUM_TIMER_WAIT, COUNT_READ, SUM_TIMER_READ, SUM_NUMBER_OF_BYTES_READ, COUNT_WRITE, SUM_TIMER_WRITE, SUM_NUMBER_OF_BYTES_WRITE, COUNT_MISC, SUM_TIMER_MISC FROM file_summary_by_instance"
 330
 331         rows, err := dbh.Query(sql)
 332         if err != nil {
 333                 log.Fatal(err)
 334         }
 335         defer rows.Close()
 336
 337         for rows.Next() {
 338                 var r file_summary_by_instance_row
 339
 340                 if err := rows.Scan(&r.FILE_NAME, &r.COUNT_STAR, &r.SUM_TIMER_WAIT, &r.COUNT_READ, &r.SUM_TIMER_READ, &r.SUM_NUMBER_OF_BYTES_READ, &r.COUNT_WRITE, &r.SUM_TIMER_WRITE, &r.SUM_NUMBER_OF_BYTES_WRITE, &r.COUNT_MISC, &r.SUM_TIMER_MISC); err != nil {
 341                         log.Fatal(err)
 342                 }
 343                 t = append(t, r)
 344         }
 345         if err := rows.Err(); err != nil {
 346                 log.Fatal(err)
 347         }
 348         lib.Logger.Println("select_fsbi_rows() took:", time.Duration(time.Since(start)).String())
 349
 350         return t
 351 }
 352
 353 // remove the initial values from those rows where there's a match
 354 // - if we find a row we can't match ignore it
 355 func (this *file_summary_by_instance_rows) subtract(initial file_summary_by_instance_rows) {
 356         i_by_name := make(map[string]int)
 357
 358         // iterate over rows by name
 359         for i := range initial {
 360                 i_by_name[initial[i].name()] = i
 361         }
 362
 363         for i := range *this {
 364                 if _, ok := i_by_name[(*this)[i].name()]; ok {
 365                         initial_i := i_by_name[(*this)[i].name()]
 366                         (*this)[i].subtract(initial[initial_i])
 367                 }
 368         }
 369 }
 370
 371 func (t file_summary_by_instance_rows) Len() int      { return len(t) }
 372 func (t file_summary_by_instance_rows) Swap(i, j int) { t[i], t[j] = t[j], t[i] }
 373 func (t file_summary_by_instance_rows) Less(i, j int) bool {
 374         return (t[i].SUM_TIMER_WAIT > t[j].SUM_TIMER_WAIT) ||
 375                 ((t[i].SUM_TIMER_WAIT == t[j].SUM_TIMER_WAIT) && (t[i].FILE_NAME < t[j].FILE_NAME))
 376 }
 377
 378 func (t *file_summary_by_instance_rows) sort() {
 379         sort.Sort(t)
 380 }
 381
 382 // if the data in t2 is "newer", "has more values" than t then it needs refreshing.
 383 // check this by comparing totals.
 384 func (t file_summary_by_instance_rows) needs_refresh(t2 file_summary_by_instance_rows) bool {
 385         my_totals := t.totals()
 386         t2_totals := t2.totals()
 387
 388         return my_totals.SUM_TIMER_WAIT > t2_totals.SUM_TIMER_WAIT
 389 }