Incremental Checkpoints

Question

Hi,
I have a Oracle 10.2.0.4 Standard Edition database with a manual physical standby on a remote server. To synchronize my stnadby, I'm forcing a log switch then I copy the archive log on the standby site for recovery.
I'm having several "Checkpoint not complete" in the alert log and I found the reason: when you switch the log, there's only an incremental checkpoint not a full one... so when I force the switch and the system wants to recycle the first log (the one where the last full checkpoint occurs), I'm getting the warning because the database must complete the checkpoint.
My question: Is performing a "ALTER SYSTEM CHECKPOINT" before my log switch the solution to my problem or is there a better way to do so?
Thanks
Christophe

Answer 1

Hi Pavan,
I don't think there's any link with slow IO: as I said, the system is doing nothing and the archived logs are from 500K to 1.2M, with my redologs being at 50M.
For the FAST_START_MTTR_TARGET, I tested it with a value of 30 secondes and it does complete incremental checkpoint.
Here's an example of what is happening:
SQL> alter system checkpoint;
System altered.
SQL> select * from v$sysstat where name like '%background checkpoint%';
STATISTIC# NAME                                                                  CLASS      VALUE
       166 background checkpoints started                                            8          2
       167 background checkpoints completed                                          8          2
SQL> alter system switch logfile;
System altered.
SQL> select * from v$sysstat where name like '%background checkpoint%';
STATISTIC# NAME                                                                  CLASS      VALUE
       166 background checkpoints started                                            8          3
       167 background checkpoints completed                                          8          2
SQL> alter system switch logfile;
System altered.
SQL> select * from v$sysstat where name like '%background checkpoint%';
STATISTIC# NAME                                                                  CLASS      VALUE
       166 background checkpoints started                                            8          4
       167 background checkpoints completed                                          8          2
SQL> alter system checkpoint;
System altered.
SQL> select * from v$sysstat where name like '%background checkpoint%';
STATISTIC# NAME                                                                  CLASS      VALUE
       166 background checkpoints started                                            8          5
       167 background checkpoints completed                                          8          5I think, that in my case, the only option is to force a checkpoint before I force the switch to avoid the system to issue a "Checkpoint not complete" with all the transactions stopped while it complete the checkpoint.
Thanks
Christophe