前言

前幾天在處理一個生產環境的數據庫恢復任務，2TB 的 Oracle 數據庫需要從備份中完整恢復。按照慣例，晚上啟動 RMAN 恢復任務，準備讓它跑一整夜。

結果第二天早上一看，磁盤空間不夠，RMAN 恢復中斷了。怎么辦，重來嗎？

本文記錄一下處理過程，方便后續查看。

問題現象

查看 RMAN 恢復日志：

channel c2: reading from backup piece /backup/db_full_xxx_1_1_1_20250825.bak
channel c1: ORA-19870: error while restoring backup piece /data/backup/db_full_xxx_1_1_1_20250825.bak
ORA-19502: write error on file "/data/database/tbs_data04.dbf", block number 3145984 (block size=8192)
ORA-27072: File I/O error
Linux-x86_64 Error: 28: No space left on device
Additional information: 4
Additional information: 3145984
Additional information: -1

channel c2: ORA-19870: error while restoring backup piece /data/backup/db_full_xxx_1_1_1_20250825.bak
ORA-19502: write error on file "/data/database/tbs_data.327.1170146731", block number 37184 (block size=8192)
ORA-27072: File I/O error
Additional information: 4
Additional information: 37184
Additional information: 217088

channel c3: ORA-19870: error while restoring backup piece /data/backup/db_full_xxx_1_1_1_20250825.bak
ORA-19502: write error on file "/data/database/tbs_data.311.1147251363", block number 3219328 (block size=8192)
ORA-27072: File I/O error
Linux-x86_64 Error: 28: No space left on device

failover to previous backup

creating datafile file number=16 name=/data/database/tbs_data.dbf
creating datafile file number=36 name=/data/database/tbs_data04.dbf
creating datafile file number=37 name=/data/database/tbs_data05.dbf
[部分數據文件創建日志省略...]
Finished restore at 02-SEP-25

RMAN-00571: ===========================================================
RMAN-00569: =============== ERROR MESSAGE STACK FOLLOWS ===============
RMAN-00571: ===========================================================
RMAN-03002: failure of switch command at 09/02/2025 21:19:55
ORA-19563: header validation failed for file

Recovery Manager complete.

從報錯 Linux-x86_64 Error: 28: No space left on device 明顯可以看到是磁盤空間不足導致恢復中斷。

磁盤擴容

首先需要解決磁盤空間問題，虛擬化增加了一塊 2T 盤，執行以下 LVM 擴容操作：

# 創建物理卷
pvcreate /dev/sdc
# 擴展卷組
vgextend data /dev/sdc
# 擴展邏輯卷（使用全部可用空間）
lvextend -l +100%FREE /dev/data/data_lv
# 調整文件系統大小
resize2fs /dev/data/data_lv

擴容完成后，確認磁盤空間充足后準備重新啟動恢復任務。

RMAN 斷點續傳機制

一個關鍵問題：RMAN 是否支持從中斷點繼續恢復，還是需要重新開始？

啟動恢復腳本進行驗證：

[oracle@dbserver:/home/oracle]$ sh restore.sh &
[1] 16640
[oracle@dbserver:/home/oracle]$ tail -200f restore_2025-09-03.log

觀察日志輸出，可以清楚看到 RMAN 的智能恢復機制：

Starting restore at 03-SEP-25

datafile 1 is already restored to file /data/database/system.260.1056479481
datafile 2 is already restored to file /data/database/sysaux.261.1056479483
datafile 3 is already restored to file /data/database/undotbs1.262.1056479483
datafile 4 is already restored to file /data/database/undotbs2.264.1056479487
datafile 5 is already restored to file /data/database/users.265.1056479487
datafile 6 is already restored to file /data/database/tbs_trans
datafile 7 is already restored to file /data/database/tbs_workflow_ex
datafile 8 is already restored to file /data/database/tbs_hist
datafile 9 is already restored to file /data/database/tbs_main
[已恢復數據文件列表省略...]

# 從未完成的數據文件開始恢復
channel c1: starting datafile backup set restore
channel c1: specifying datafile(s) to restore from backup set
channel c1: restoring datafile 00036 to /data/database/tbs_data04.dbf
channel c1: restoring datafile 00046 to /data/database/tbs_data.310.1147251361
channel c1: restoring datafile 00061 to /data/database/tbs_data.325.1165308233
channel c1: restoring datafile 00083 to /data/database/tbs_data.347.1189929603
channel c1: reading from backup piece /backup/db_full_xxx_1_1_1_20250901.bak

監控恢復進度

使用以下 SQL 查詢實時監控恢復進度：

SQL> set line2222 pages1000
SQL> SELECT SID, SERIAL#, CONTEXT, SOFAR, TOTALWORK,
       ROUND(SOFAR/TOTALWORK*100,2) "% Complete",
       TIME_REMAINING/60 "Minutes Remaining"
FROM V$SESSION_LONGOPS
WHERE OPNAME LIKE 'RMAN%'
  AND TOTALWORK != 0
  AND SOFAR != TOTALWORK; 

       SID    SERIAL#    CONTEXT      SOFAR  TOTALWORK % Complete Minutes Remaining
---------- ---------- ---------- ---------- ---------- ---------- -----------------
      1423         14          1    1063783   15990656       6.65        66.4166667
      1992          7          1     809302   12312064       6.57        67.2833333
       853        607          3    2569277  262981344        .98        413.866667
      1707          7          1    1091712   15990656       6.83              64.6

寫在最后

通過這個案例，想跟大家分享一個很多人可能不太了解的 RMAN 特性。很多 DBA 朋友遇到恢復中斷時，第一反應是重新開始，其實完全沒必要。

RMAN 的斷點續傳機制相當可靠，這個功能從 Oracle 9i 開始就有了，但真正用過的人不多。原理很簡單：RMAN 會檢查目標文件的 header 信息和備份集的元數據，判斷哪些數據文件已經完整恢復。

對于剛接觸 Oracle 的朋友，遇到大型數據庫恢復中斷不要慌。RMAN 比你想象中聰明，相信它的斷點續傳能力。這個機制在企業級應用中經過了無數次驗證，穩定性和可靠性都沒問題。

當然，預防永遠比補救重要。恢復前的規劃和檢查，遠比事后的補救措施來得有效。