Tuning Recovery - Expert Oracle RAC Performance Diagnostics and Tuning - page 319

13. While cache recovery takes place, work occurs at a slower pace. Once this stage of the

recovery operation is complete, it is considered full database availability, now all data is
accessible including that which resided on the failed instance.

14. After completion of the cache recovery process, Oracle starts the transaction recovery

operation, that is, rolling back of all uncommitted transactions.

• Compared to cache recovery where the recovery is of a forward nature, that is, rolling

forward of the transactions from the redo logs, the transaction recovery handles
uncommitted transactions; hence, operation is to roll back all uncommitted transactions
of a failed instance. In addition, during this pass, the redo threads for the failed instances
are merged by SCN, and the redo is applied to the data files.

• During this process of rolling back uncommitted transactions, Oracle uses a technology

called fast-start fault recovery in which it performs the transaction recovery as a deferred
process as a background activity. Oracle uses a multi-version and consistency method
to provide on-demand rollback of only those rows blocked by expired transactions. This
feature prevents new transactions from waiting for the rollback activity to complete.

• Depending on how the transaction failed, the recovery is performed either by the

SMON process or by the server process that initiated the transaction. The following
output illustrates the various steps from the time the redo log is scanned to the activity
performed by the SMON process completing the recovery.

Wed Aug 27 10:07:59 2014
Completed redo scan
read 3298 KB redo, 589 data blocks need recovery
* validated domain 0, flags = 0x0

Wed Aug 27 10:08:00 2014
Started redo application at
Thread 2: logseq 353, block 1930

Wed Aug 27 10:08:03 2014
Recovery of Online Redo Log: Thread 2 Group 3 Seq 353 Reading mem 0
Mem# 0: +SSKY_DATA/SSKYDB/ONLINELOG/group_3.266.854233685
Mem# 1: +SSKY_FRA/SSKYDB/ONLINELOG/group_3.259.854233687

Wed Aug 27 10:08:04 2014
Completed redo application of 1.93MB
Wed Aug 27 10:08:10 2014
Completed instance recovery at
Thread 2: logseq 353, block 8527, scn 36904445
498 data blocks read, 765 data blocks written, 3298 redo k-bytes read

Wed Aug 27 10:08:13 2014
Thread 2 advanced to log sequence 354 (thread recovery)
SMON[INST-TXN-RECO]:about to recover undo segment 1 status:3 inst:1
SMON[INST-TXN-RECO]:about to recover undo segment 2 status:3 inst:1
SMON[INST-TXN-RECO]:about to recover undo segment 3 status:3 inst:1
SMON[INST-TXN-RECO]:about to recover undo segment 4 status:3 inst:1
SMON[INST-TXN-RECO]:about to recover undo segment 5 status:3 inst:1
SMON[INST-TXN-RECO]:about to recover undo segment 6 status:3 inst:1