performance_schema/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         //      "strconv"
  12         "time"
  13
  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 )
  71
  72 type file_summary_by_instance_row struct {
  73         FILE_NAME string
  74
  75         COUNT_STAR  uint64
  76         COUNT_READ  uint64
  77         COUNT_WRITE uint64
  78         COUNT_MISC  uint64
  79
  80         SUM_TIMER_WAIT  uint64
  81         SUM_TIMER_READ  uint64
  82         SUM_TIMER_WRITE uint64
  83         SUM_TIMER_MISC  uint64
  84
  85         SUM_NUMBER_OF_BYTES_READ  uint64
  86         SUM_NUMBER_OF_BYTES_WRITE uint64
  87 }
  88
  89 // represents a table or set of rows
  90 type file_summary_by_instance_rows []file_summary_by_instance_row
  91
  92 // Return the name using the FILE_NAME attribute.
  93 func (r *file_summary_by_instance_row) name() string {
  94         return r.FILE_NAME
  95 }
  96
  97 // Return a formatted pretty name for the row.
  98 func (r *file_summary_by_instance_row) pretty_name() string {
  99         s := r.name()
 100         if len(s) > 30 {
 101                 s = s[:29]
 102         }
 103         return fmt.Sprintf("%-30s", s)
 104 }
 105
 106 func (r *file_summary_by_instance_row) headings() string {
 107         return fmt.Sprintf("%-30s %10s %6s|%6s %6s %6s|%8s %8s|%8s %6s %6s %6s",
 108                 "Table Name",
 109                 "Latency",
 110                 "%",
 111                 "Read",
 112                 "Write",
 113                 "Misc",
 114                 "Rd bytes",
 115                 "Wr bytes",
 116                 "Ops",
 117                 "R Ops",
 118                 "W Ops",
 119                 "M Ops")
 120 }
 121
 122 // generate a printable result
 123 func (row *file_summary_by_instance_row) row_content(totals file_summary_by_instance_row) string {
 124         var name string
 125
 126         // We assume that if COUNT_STAR = 0 then there's no data at all...
 127         // when we have no data we really don't want to show the name either.
 128         if row.COUNT_STAR == 0 {
 129                 name = ""
 130         } else {
 131                 name = row.pretty_name()
 132         }
 133
 134         return fmt.Sprintf("%-30s %10s %6s|%6s %6s %6s|%8s %8s|%8s %6s %6s %6s",
 135                 name,
 136                 lib.FormatTime(row.SUM_TIMER_WAIT),
 137                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_WAIT, totals.SUM_TIMER_WAIT)),
 138                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_READ, row.SUM_TIMER_WAIT)),
 139                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_WRITE, row.SUM_TIMER_WAIT)),
 140                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_MISC, row.SUM_TIMER_WAIT)),
 141                 lib.FormatAmount(row.SUM_NUMBER_OF_BYTES_READ),
 142                 lib.FormatAmount(row.SUM_NUMBER_OF_BYTES_WRITE),
 143                 lib.FormatAmount(row.COUNT_STAR),
 144                 lib.FormatPct(lib.MyDivide(row.COUNT_READ, row.COUNT_STAR)),
 145                 lib.FormatPct(lib.MyDivide(row.COUNT_WRITE, row.COUNT_STAR)),
 146                 lib.FormatPct(lib.MyDivide(row.COUNT_MISC, row.COUNT_STAR)))
 147 }
 148
 149 func (this *file_summary_by_instance_row) add(other file_summary_by_instance_row) {
 150         this.COUNT_STAR += other.COUNT_STAR
 151         this.COUNT_READ += other.COUNT_READ
 152         this.COUNT_WRITE += other.COUNT_WRITE
 153         this.COUNT_MISC += other.COUNT_MISC
 154
 155         this.SUM_TIMER_WAIT += other.SUM_TIMER_WAIT
 156         this.SUM_TIMER_READ += other.SUM_TIMER_READ
 157         this.SUM_TIMER_WRITE += other.SUM_TIMER_WRITE
 158         this.SUM_TIMER_MISC += other.SUM_TIMER_MISC
 159
 160         this.SUM_NUMBER_OF_BYTES_READ += other.SUM_NUMBER_OF_BYTES_READ
 161         this.SUM_NUMBER_OF_BYTES_WRITE += other.SUM_NUMBER_OF_BYTES_WRITE
 162 }
 163
 164 func (this *file_summary_by_instance_row) subtract(other file_summary_by_instance_row) {
 165         this.COUNT_STAR -= other.COUNT_STAR
 166         this.COUNT_READ -= other.COUNT_READ
 167         this.COUNT_WRITE -= other.COUNT_WRITE
 168         this.COUNT_MISC -= other.COUNT_MISC
 169
 170         this.SUM_TIMER_WAIT -= other.SUM_TIMER_WAIT
 171         this.SUM_TIMER_READ -= other.SUM_TIMER_READ
 172         this.SUM_TIMER_WRITE -= other.SUM_TIMER_WRITE
 173         this.SUM_TIMER_MISC -= other.SUM_TIMER_MISC
 174
 175         this.SUM_NUMBER_OF_BYTES_READ -= other.SUM_NUMBER_OF_BYTES_READ
 176         this.SUM_NUMBER_OF_BYTES_WRITE -= other.SUM_NUMBER_OF_BYTES_WRITE
 177 }
 178
 179 // return the totals of a slice of rows
 180 func (t file_summary_by_instance_rows) totals() file_summary_by_instance_row {
 181         var totals file_summary_by_instance_row
 182         totals.FILE_NAME = "TOTALS"
 183
 184         for i := range t {
 185                 totals.add(t[i])
 186         }
 187
 188         return totals
 189 }
 190
 191 // clean up the given path reducing redundant stuff and return the clean path
 192 func cleanup_path(path string) string {
 193
 194         for {
 195                 orig_path := path
 196                 path = re_one_or_the_other.ReplaceAllString(path, "/")
 197                 path = re_slash_dot_dot_slash.ReplaceAllString(path, "/")
 198                 if orig_path == path { // no change so give up
 199                         break
 200                 }
 201         }
 202
 203         return path
 204 }
 205
 206 // From the original FILE_NAME we want to generate a simpler name to use.
 207 // This simpler name may also merge several different filenames into one.
 208 func (t file_summary_by_instance_row) simple_name(global_variables map[string]string) string {
 209
 210         path := t.FILE_NAME
 211
 212         if cached_result, err := get_from_cache(path); err == nil {
 213                 return cached_result
 214         }
 215
 216         // FIXME and make this work.
 217         //      re4 := regexp.MustCompile(re_encoded)
 218         //      if m4 := re4.FindStringSubmatch(path); m4 != nil {
 219         //              if value, err := strconv.ParseInt(m4[1], 16, 16); err != nil {
 220         //                      // missing replace @.... with char(value) in path
 221         //
 222         //              }
 223         //      }
 224
 225         // this should probably be ordered from most expected regexp to least
 226         if m1 := re_table_file.FindStringSubmatch(path); m1 != nil {
 227                 // we may match temporary tables so check for them
 228                 if m2 := re_temp_table.FindStringSubmatch(m1[2]); m2 != nil {
 229                         return save_to_cache(path, "<temp_table>")
 230                 }
 231
 232                 // we may match partitioned tables so check for them
 233                 if m3 := re_part_table.FindStringSubmatch(m1[2]); m3 != nil {
 234                         return save_to_cache(path, m1[1]+"."+m3[1]) // <schema>.<table> (less partition info)
 235                 }
 236
 237                 return save_to_cache(path, m1[1]+"."+m1[2]) // <schema>.<table>
 238         }
 239         if re_ibdata.MatchString(path) == true {
 240                 return save_to_cache(path, "<ibdata>")
 241         }
 242         if re_redo_log.MatchString(path) == true {
 243                 return save_to_cache(path, "<redo_log>")
 244         }
 245         if re_binlog.MatchString(path) == true {
 246                 return save_to_cache(path, "<binlog>")
 247         }
 248         if re_db_opt.MatchString(path) == true {
 249                 return save_to_cache(path, "<db_opt>")
 250         }
 251         if re_slowlog.MatchString(path) == true {
 252                 return save_to_cache(path, "<slow_log>")
 253         }
 254         if re_auto_cnf.MatchString(path) == true {
 255                 return save_to_cache(path, "<auto_cnf>")
 256         }
 257         // relay logs are a bit complicated. If a full path then easy to
 258         // identify,but if a relative path we may need to add $datadir,
 259         // but also if as I do we have a ../blah/somewhere/path then we
 260         // need to make it match too.
 261         if len(global_variables["relay_log"]) > 0 {
 262                 relay_log := global_variables["relay_log"]
 263                 if relay_log[0] != '/' { // relative path
 264                         relay_log = cleanup_path(global_variables["datadir"] + relay_log) // datadir always ends in /
 265                 }
 266                 re_relay_log := relay_log + `\.(\d{6}|index)$`
 267                 if regexp.MustCompile(re_relay_log).MatchString(path) == true {
 268                         return save_to_cache(path, "<relay_log>")
 269                 }
 270         }
 271         if re_pid_file.MatchString(path) == true {
 272                 return save_to_cache(path, "<pid_file>")
 273         }
 274         if re_error_msg.MatchString(path) == true {
 275                 return save_to_cache(path, "<errmsg>")
 276         }
 277         if re_charset.MatchString(path) == true {
 278                 return save_to_cache(path, "<charset>")
 279         }
 280         return save_to_cache(path, path)
 281 }
 282
 283 // Convert the imported "table" to a merged one with merged data.
 284 // Combine all entries with the same "FILE_NAME" by adding their values.
 285 func merge_by_table_name(orig file_summary_by_instance_rows, global_variables map[string]string) file_summary_by_instance_rows {
 286         start := time.Now()
 287         t := make(file_summary_by_instance_rows, 0, len(orig))
 288
 289         m := make(map[string]file_summary_by_instance_row)
 290
 291         // iterate over source table
 292         for i := range orig {
 293                 var file_name string
 294                 var new_row file_summary_by_instance_row
 295                 orig_row := orig[i]
 296
 297                 if orig_row.COUNT_STAR > 0 {
 298                         file_name = orig_row.simple_name(global_variables)
 299
 300                         // check if we have an entry in the map
 301                         if _, found := m[file_name]; found {
 302                                 new_row = m[file_name]
 303                         } else {
 304                                 new_row.FILE_NAME = file_name
 305                         }
 306                         new_row.add(orig_row)
 307                         m[file_name] = new_row // update the map with the new value
 308                 }
 309         }
 310
 311         // add the map contents back into the table
 312         for _, row := range m {
 313                 t = append(t, row)
 314         }
 315
 316         lib.Logger.Println("merge_by_table_name() took:", time.Duration(time.Since(start)).String())
 317         return t
 318 }
 319
 320 // Select the raw data from the database into file_summary_by_instance_rows
 321 // - filter out empty values
 322 // - merge rows with the same name into a single row
 323 // - change FILE_NAME into a more descriptive value.
 324 func select_fsbi_rows(dbh *sql.DB) file_summary_by_instance_rows {
 325         var t file_summary_by_instance_rows
 326         start := time.Now()
 327
 328         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"
 329
 330         rows, err := dbh.Query(sql)
 331         if err != nil {
 332                 log.Fatal(err)
 333         }
 334         defer rows.Close()
 335
 336         for rows.Next() {
 337                 var r file_summary_by_instance_row
 338
 339                 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 {
 340                         log.Fatal(err)
 341                 }
 342                 t = append(t, r)
 343         }
 344         if err := rows.Err(); err != nil {
 345                 log.Fatal(err)
 346         }
 347         lib.Logger.Println("select_fsbi_rows() took:", time.Duration(time.Since(start)).String())
 348
 349         return t
 350 }
 351
 352 // remove the initial values from those rows where there's a match
 353 // - if we find a row we can't match ignore it
 354 func (this *file_summary_by_instance_rows) subtract(initial file_summary_by_instance_rows) {
 355         i_by_name := make(map[string]int)
 356
 357         // iterate over rows by name
 358         for i := range initial {
 359                 i_by_name[initial[i].name()] = i
 360         }
 361
 362         for i := range *this {
 363                 if _, ok := i_by_name[(*this)[i].name()]; ok {
 364                         initial_i := i_by_name[(*this)[i].name()]
 365                         (*this)[i].subtract(initial[initial_i])
 366                 }
 367         }
 368 }
 369
 370 func (t file_summary_by_instance_rows) Len() int      { return len(t) }
 371 func (t file_summary_by_instance_rows) Swap(i, j int) { t[i], t[j] = t[j], t[i] }
 372 func (t file_summary_by_instance_rows) Less(i, j int) bool {
 373         return (t[i].SUM_TIMER_WAIT > t[j].SUM_TIMER_WAIT) ||
 374                 ((t[i].SUM_TIMER_WAIT == t[j].SUM_TIMER_WAIT) && (t[i].FILE_NAME < t[j].FILE_NAME))
 375 }
 376
 377 func (t *file_summary_by_instance_rows) sort() {
 378         sort.Sort(t)
 379 }
 380
 381 // if the data in t2 is "newer", "has more values" than t then it needs refreshing.
 382 // check this by comparing totals.
 383 func (t file_summary_by_instance_rows) needs_refresh(t2 file_summary_by_instance_rows) bool {
 384         my_totals := t.totals()
 385         t2_totals := t2.totals()
 386
 387         return my_totals.SUM_TIMER_WAIT > t2_totals.SUM_TIMER_WAIT
 388 }