PostgreSQL資料庫伺服器在響應客戶端發出備份命令pg_basebackup時,到底做了什麼?本節一探究竟.

### 一、資料結構
**basebackup_options**
pg_basebackup的選項,在資料庫伺服器解析為該資料結構.
```
typedef struct
{
  //備份的標籤
    const char *label;
    //是否顯示進度
    bool        progress;
    //是否執行快速fast checkpoint?
    bool        fastcheckpoint;
    //nowait?
    bool        nowait;
    //是否包含wal data
    bool        includewal;
    //
    uint32      maxrate;
    //是否包含表空間對映檔案?
    bool        sendtblspcmapfile;
} basebackup_options;

```

### 二、原始碼解讀
資料庫伺服器接收到請求,postmaster啟動新的postgres程式響應此請求,此程式被視為walsender,標記am_walsender設定為T,在PostgresMain函式中,將執行以下邏輯:
```
...
  for (;;)//主迴圈
  {
    ...
    switch (firstchar)
    {
      case 'Q':     /* simple query */
        {
          if (am_walsender)
          {
            //如為WAL sender,執行exec_replication_command
            if (!exec_replication_command(query_string))
              exec_simple_query(query_string);
            ...
...
```
呼叫exec_replication_command函式,執行相關命令.該函式會呼叫SendBaseBackup函式執行具體的實現邏輯,其中重點的實現函式是sendFileWithContent/sendDir.
1.sendFileWithContent函式用於傳送backup_label等檔案到客戶端
pq_putmessage傳送訊息,'d'的訊息型別表示CopyData.
```
static void
sendFileWithContent(const char *filename, const char *content)
{
    struct stat statbuf;
    int         pad,
                len;

len = strlen(content);

/*
     * Construct a stat struct for the backup_label file we're injecting in
     * the tar.
     */
    /* Windows doesn't have the concept of uid and gid */
#ifdef WIN32
    statbuf.st_uid = 0;
    statbuf.st_gid = 0;
#else
    statbuf.st_uid = geteuid();
    statbuf.st_gid = getegid();
#endif
    statbuf.st_mtime = time(NULL);
    statbuf.st_mode = pg_file_create_mode;
    statbuf.st_size = len;

_tarWriteHeader(filename, NULL, &statbuf, false);
    /* Send the contents as a CopyData message */
    pq_putmessage('d', content, len);

/* Pad to 512 byte boundary, per tar format requirements */
    pad = ((len + 511) & ~511) - len;
    if (pad > 0)
    {
        char        buf[512];

MemSet(buf, 0, pad);
        pq_putmessage('d', buf, pad);
    }
}
```
2.sendDir遍歷檔案目錄,呼叫sendFile傳送到客戶端
遞迴遍歷資料庫目錄,呼叫sendFile傳送檔案
```
...

if (!sizeonly)
                sent = sendFile(pathbuf, pathbuf + basepathlen + 1, &statbuf,
                                true, isDbDir ? pg_atoi(lastDir + 1, sizeof(Oid), 0) : InvalidOid);

if (sent || sizeonly)
            {
                /* Add size, rounded up to 512byte block */
                size += ((statbuf.st_size + 511) & ~511);
                size += 512;    /* Size of the header of the file */
            }
...            
```
sendFile傳送相應的檔案內容到客戶端
```
...
    while ((cnt = fread(buf, 1, Min(sizeof(buf), statbuf->st_size - len), fp)) > 0)
    {
      ...
       /* Send the chunk as a CopyData message */
        if (pq_putmessage('d', buf, cnt))
            ereport(ERROR,
                    (errmsg("base backup could not send data, aborting backup")));
        ...
    }

```

### 三、跟蹤分析
客戶端啟動pg_basebackup
```
[xdb@localhost ~]$ pg_basebackup -h 192.168.26.25 -U replicator -p 5432 -D /data/backup -P -Xs -R
Password:

```

跟蹤postmaster,設定跟蹤子程式
```
(gdb) set follow-fork-mode child
(gdb) b PostgresMain

```
客戶端輸入密碼後,進入斷點,在執行BASE_BACKUP命令前,首先會執行SHOW data_directory_mode/SHOW wal_segment_size/IDENTIFY_SYSTEM三個命令,然後再執行BASE_BACKUP命令
```
...
(gdb) p input_message
$2 = {data = 0x20a1d78 "SHOW data_directory_mode", len = 25, maxlen = 1024, cursor = 0}
...
(gdb) p input_message
$4 = {data = 0x20a1d78 "SHOW wal_segment_size", len = 22, maxlen = 1024, cursor = 0}
...
(gdb) p input_message
$5 = {data = 0x20a1d78 "IDENTIFY_SYSTEM", len = 16, maxlen = 1024, cursor = 0}
...
(gdb) p input_message
$7 = {data = 0x20a1d78 "BASE_BACKUP LABEL 'pg_basebackup base backup' PROGRESS   NOWAIT   ", len = 67, maxlen = 1024, 
  cursor = 0}
...
```
跟蹤SendBaseBackup
```
(gdb) b SendBaseBackup
...
4178              if (!exec_replication_command(query_string))
(gdb) step
exec_replication_command (cmd_string=0x20a1d78 "BASE_BACKUP LABEL 'pg_basebackup base backup' PROGRESS   NOWAIT   ")
    at walsender.c:1438
1438    if (got_STOPPING)

1521        SendBaseBackup((BaseBackupCmd *) cmd_node);
...
```
進入SendBaseBackup
```
(gdb) c
Continuing.
[New process 1811]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
[Switching to Thread 0x7f46389b58c0 (LWP 1811)]

Breakpoint 1, SendBaseBackup (cmd=0x2137da8) at basebackup.c:762
762   parse_basebackup_options(cmd->options, &opt);
(gdb) n
764   WalSndSetState(WALSNDSTATE_BACKUP);
(gdb) p opt
$1 = {label = 0x2137760 "pg_basebackup base backup", progress = true, fastcheckpoint = false, nowait = true, 
  includewal = false, maxrate = 0, sendtblspcmapfile = false}
```
實際執行backup的是函式perform_base_backup
```
...
(gdb) 
775   perform_base_backup(&opt);
(gdb) step
perform_base_backup (opt=0x7ffc96573180) at basebackup.c:232
```
執行sendXXX
```
320       if (ti->path == NULL)
(gdb) 
325         sendFileWithContent(BACKUP_LABEL_FILE, labelfile->data);
(gdb) p *labelfile
$4 = {
  data = 0x2138a50 "START WAL LOCATION: 0/66000028 (file 0000001", '0' <repeats times="">, "66)\nCHECKPOINT LOCATION: 0/66000060\nBACKUP METHOD: streamed\nBACKUP FROM: master\nSTART TIME: 2019-03-26 17:05:45 CST\nLABEL: pg_basebackup base"..., 
  len = 227, maxlen = 1024, cursor = 0}
(gdb) n
331         if (tblspc_map_file && opt->sendtblspcmapfile)
(gdb) 
337           sendDir(".", 1, false, tablespaces, true);
(gdb) 
340         if (lstat(XLOG_CONTROL_FILE, &statbuf) != 0)
(gdb) 
345         sendFile(XLOG_CONTROL_FILE, XLOG_CONTROL_FILE, &statbuf, false);
(gdb) 
356       if (opt->includewal && ti->path == NULL)
(gdb) 
361         pq_putemptymessage('c');  /* CopyDone */
(gdb) 
309     foreach(lc, tablespaces)
(gdb) 
364     endptr = do_pg_stop_backup(labelfile->data, !opt->nowait, &endtli);
(gdb) 
366   PG_END_ENSURE_ERROR_CLEANUP(base_backup_cleanup, (Datum) 0);
(gdb) 
369   if (opt->includewal)
(gdb) 
605   SendXlogRecPtrResult(endptr, endtli);
(gdb) 
607   if (total_checksum_failures)
(gdb) 
623 }
(gdb) 
SendBaseBackup (cmd=0x2137da8) at basebackup.c:776
776 }
(gdb) 
exec_replication_command (cmd_string=0x20a1d78 "BASE_BACKUP LABEL 'pg_basebackup base backup' PROGRESS   NOWAIT   ")
    at walsender.c:1522
1522        break;
(gdb) c
Continuing.
[Inferior 2 (process 1811) exited normally]
(gdb)

```
DONE!

### 四、參考資料
[PG Source Code](https://doxygen.postgresql.org)</repeats>

PostgreSQL 原始碼解讀（158）-PG Tools#5(pg_basebackup：what does db server do)

相關文章