| Checkpoints and Instance recovery [message #369345] |
Mon, 21 August 2000 12:40 |
tinku
Messages: 5
Registered: August 2000
|
Junior Member |
|
|
I perhaps have misunderstood the details and hence have the following confusion:
During instance recovery all online redo logfiles written since the last checkpoint is required.
Suppose the redo log files are 1GB each and settings are such that checkpoints occur only at log switch. Because of transactions going on today till 5 p.m., the active redo log file gets filled to say 500MB. Then no transactions occur and I take a full hot backup tonight. Tomorrow morning transaction starts and db crashes before another log switch.
I recover from the nights backup. Now the datafiles already have the changes corresponding to first 500M of the redo log file (since db was idle after 5 p.m. and DBWR writes after every 3 seconds). So Oracle should not apply these changes from the redo log file to the data file - How is this managed.
Further during checkpoints is it ensured that the redo log entries are written first and then the corresponding dirty data buffers are written on to the datafiles.
Thanks in advance.
Tinku
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
