May someone help me to know when capture/pump/replicat process does a checkpoint? There are 2 things which made me confused.
- CHECKPOINTSECS parameter: There is no setting in the parameter file, that means it will be set by default (10s)
- One document says that checkpoints are based on transaction boundaries, that means checkpoint will happened by the end of each transaction
In my source db, there was a long running transaction lasting for 1 hour and already committed and written to the local trail files. Then the pump process have started to read the trail files to pump to target. My concern is when the capture/pump/replicat process do a checkpoint? 10s or 1 hour?
One thing is I often see the value of "Time since chkpt" in the output of "Info all" command is more than 10s. If they do a checkpoint every 10s, how could the value of "Time since chkpt" be more than 10s? I think that maybe the term "chkpt" of "Time since chkpt" is of database transaction, not of GoldenGate. However I'm not sure. It made me really confused.
Here is the output of "Info all" command:
Program Status Group Lag Time Since Chkpt
EXTRACT RUNNING PUMP6 00:17:07 00:04:29
My system profile:
OS: OEL5 64bit
DB: Oracle 11gR2 (184.108.40.206)
GG: 11gR1 (11.1)
Thanks for your comment. I understand that there are some kinds of checkpoint for each process. However I don't know exactly which parameter is to control the checkpoint interval of each kind. Could you elaborate more about it, either for extract, pump or replicat process?
CHECKPOINTSECS defines how often an extract makes its regular checkpoints. An extract will make a checkpoint every 10 seconds if this value is set to 10 irrespective of a transaction that is running for 1 hour. Extract also writes commit checkpoints when a transaction is committed.
In the example where your Pump is showing time since checkpoint as greater than 10 could have a problem at the network level or issues with the remote trail..
that is my observation as I had an extract that was being shown as RUNNING but time since last checkpoint was in hours and when i checked the report file, it had an error.