Document Type | Technical Information
Category | Backup/Recovery
Applicable Product Versions | 6FS07PS, 7FS02PS
Document Number | TBATI023
Overview
The Tibero database checks the files necessary for database operation at each boot stage during restart and automatically performs recovery by boot stage if recovery is required.
Tibero's recovery operations include Crash Recovery, which automatically recovers using Checkpoint and TSN (Transaction Sequence Number) during restart after abnormal database shutdown, and Media Recovery, which recovers when physical damage occurs to the files composing Tibero.
This document guides the Crash Recovery procedure that automatically performs recovery during Tibero startup.
Method
Roles by Tibero Boot Mode
MODE | Description | Main Tasks |
|---|---|---|
NOMOUNT | Instance start (SGA + background processes) | Database creation, control file recreation |
MOUNT | Control file open | Verify Redo and data file information, set Archive mode, perform Media Recovery |
NORMAL | Open all data files | Compare TSN and Checkpoint status, then perform Crash Recovery |
Types of Recovery After Tibero Abnormal Shutdown
Recovery Type | Description |
|---|---|
Crash Recovery | Automatic recovery during DB restart after abnormal shutdown due to system failure |
Media Recovery | Manual recovery when files themselves are damaged due to disk damage, etc. |
Crash Recovery Operation Sequence
Crash Recovery, also called crash recovery or instance recovery, is the process of recovering using the Redo Log File during the restart phase after the database has shut down abnormally.
The Crash Recovery process is as follows.
1) Execute tbboot
2) (NOMOUNT) Check the Control File location from the $TB_SID.tip file
3) (MOUNT) Check the Checkpoint count of all data files stored in the Control File Header
4) (NORMAL) Compare checkpoint TSN of each data file header with stop TSN in the Control file
NoteIn step 4), if the TSN of the data file header and the TSN match, the system starts in NORMAL mode, and if the TSN do not match, Crash Recovery proceeds.
Detailed Crash Recovery Process
1. Cache Recovery (Rolling Forward)
- . Read the Redo Log and sequentially apply transactions changed after the last Checkpoint
- . Apply sequentially starting from the lowest transaction based on TSN
- . The DB Buffer Cache is restored to the state just before the failure
[Disk status] ββββββββββββββββββββββββββββββ β DataFile TSN = 1100 β ββββββββββββββββββββββββββββββ ββββββββββββββββββββββββββββββ β Redo Log TSN = 1101~1150 β β Recovery target ββββββββββββββββββββββββββββββ
2. Transaction Recovery (Rolling Backward)
- . Track uncommitted transactions from the Undo Dictionary
- . Roll Backward those transactions to ensure data integrity
[Query uncommitted transactions in Undo area β rollback]
Note
If recovery fails or is interrupted, Media Recovery is required.