一、問題

pg_ctl start啟動(dòng)時(shí)報(bào)錯(cuò)退出：pg_ctl：server did not start in time。超時(shí)時(shí)間是多少？從什么時(shí)候到哪個(gè)階段算超時(shí)？

二、分析：該信息打印位置，從后面代碼段do_start函數(shù)中可以看出

1、pg_ctl start調(diào)用start_postmaster啟動(dòng)PG的主進(jìn)程后，每隔0.1ms檢查一次postmaster.pid文件，是否已寫入ready/standby

2、總共會(huì)檢查600次，即從啟動(dòng)主進(jìn)程后，最多等待60s，如果沒有寫入ready/standby則打印上述日志并退出

3、默認(rèn)等待時(shí)間是60s，如果pg_ctl start -t指定等待時(shí)間，則等待時(shí)間為該指定時(shí)間

三、什么時(shí)候postmaster.pid文件寫入ready/standby

1、如果是主機(jī)不管有沒有設(shè)置hot standby

1）當(dāng)startup進(jìn)程恢復(fù)完成退出時(shí)，調(diào)用proc_exit函數(shù)向主進(jìn)程發(fā)送SIGCHLD信號(hào)并退出

2）主進(jìn)程接收到信號(hào)后，signal處理函數(shù)reaper調(diào)用AddToDataDirLockFile向postmaster.pid文件寫入ready

2、如果是備機(jī)即data目錄下有recovery.cnf文件，且設(shè)置了hot standby，在實(shí)際恢復(fù)前沒有到達(dá)一致性位置

1）startup進(jìn)程向主進(jìn)程發(fā)送PMSIGNAL_RECOVERY_STARTED信號(hào)，主進(jìn)程調(diào)用信號(hào)處理函數(shù)sigusr1_handler，將pmState=PM_RECOVERY

2）每次讀取下一個(gè)xlog前都會(huì)調(diào)用CheckRecoveryConsistency函數(shù)進(jìn)行一致性檢查：

2.1 進(jìn)入一致性狀態(tài)，starup進(jìn)程向主進(jìn)程發(fā)送PMSIGNAL_BEGIN_HOT_STANDBY信號(hào)，主進(jìn)程接收到信號(hào)后調(diào)用sigusr1_handler->AddToDataDirLockFile向postmaster.pid文件寫入ready

3、如果是備機(jī)即data目錄下有recovery.cnf文件，且設(shè)置了hot standby，在實(shí)際恢復(fù)前沒有到達(dá)一致性位置

1）startup進(jìn)程向主進(jìn)程發(fā)送PMSIGNAL_RECOVERY_STARTED信號(hào)，主進(jìn)程調(diào)用信號(hào)處理函數(shù)sigusr1_handler，將pmState=PM_RECOVERY

2）每次讀取下一個(gè)xlog前都會(huì)調(diào)用CheckRecoveryConsistency函數(shù)進(jìn)行一致性檢查。如果沒有進(jìn)入一致性狀態(tài)

3）本地日志恢復(fù)完成，切換日志源時(shí)同樣調(diào)用CheckRecoveryConsistency函數(shù)進(jìn)行一致性檢查

3.1 進(jìn)入一致性狀態(tài)，starup進(jìn)程向主進(jìn)程發(fā)送PMSIGNAL_BEGIN_HOT_STANDBY信號(hào)，主進(jìn)程接收到信號(hào)后調(diào)用sigusr1_handler->AddToDataDirLockFile向postmaster.pid文件寫入ready

4、如果是備機(jī)即data目錄下有recovery.cnf文件，且設(shè)置了hot standby，在實(shí)際恢復(fù)前到達(dá)一致性位置

1）startup進(jìn)程向主進(jìn)程發(fā)送PMSIGNAL_RECOVERY_STARTED信號(hào)，主進(jìn)程調(diào)用信號(hào)處理函數(shù)sigusr1_handler，將pmState=PM_RECOVERY

2）CheckRecoveryConsistency函數(shù)進(jìn)行一致性檢查，向主進(jìn)程發(fā)送PMSIGNAL_BEGIN_HOT_STANDBY信號(hào)，主進(jìn)程接收到信號(hào)后調(diào)用sigusr1_handler->AddToDataDirLockFile向postmaster.pid文件寫入ready

5、如果是備機(jī)即data目錄下有recovery.cnf文件，沒有設(shè)置hot standby

1）startup進(jìn)程向主進(jìn)程發(fā)送PMSIGNAL_RECOVERY_STARTED信號(hào)

2）主進(jìn)程接收到信號(hào)后，向postmaster.將pmState=PM_RECOVERY

四、代碼分析

1、pg_ctl start流程

do_start->
 pm_pid = start_postmaster();
 if (do_wait){
  print_msg(_("waiting for server to start..."));
  switch (wait_for_postmaster(pm_pid, false)){
   case POSTMASTER_READY:
    print_msg(_(" donen"));
    print_msg(_("server startedn"));
    break;
   case POSTMASTER_STILL_STARTING:
    print_msg(_(" stopped waitingn"));
    write_stderr(_("%s: server did not start in timen"), progname);
    exit(1);
    break;
   case POSTMASTER_FAILED:
    print_msg(_(" stopped waitingn"));
    write_stderr(_("%s: could not start servern" "Examine the log output.n"), progname);
    exit(1);
    break;
  }
 }else
  print_msg(_("server startingn"));