Fabian Merki wrote:
hi all
thanks for your help.
should i regularly call SYSCS_UTIL.SYSCS_CHECKPOINT_DATABASE
every day or so? what happens if the db gets a kill -9 and the checkpoint was
not done?
thanks
fabian
Hi Fabian,
As Knut commented in another reply, checkpoints are invoked on a regular
basis (based on the size of the log) by Derby. Calling
SYSCS_UTIL.SYSCS_CHECKPOINT_DATABASE does the same as is done by the
"automatic checkpoint" mechanism. Thus, depending on the workload on
your system, checkpoints may already be created multiple times every
day. Normally, it should not be necessary to manually invoke
checkpoints, but I can't guarantee that it won't have positive effects
in your system.
Checkpoints are implemented as transactions, and do this (simplified
version):
1) Stop all write operations while the transaction log that has not yet
been written to disk is flushed
2) Start a checkpoint transaction and gather some interesting info
3) Write data blocks that have been modified (by any transaction) to disk
4) Commit the checkpoint transaction
If a crash/kill -9/power loss/etc happens after a checkpoint has been
started but before it had the time to complete (commit), Derby will
understand that it did not complete and remove it's effects when the
database is rebooted. For the record, this is what happens for all
transactions that were active at the time of the crash/kill -9/...: If a
transaction did not complete, it's effects are removed at database
reboot time.
If you are interested in more details about checkpoints in Derby, I
suggest you have a look at the following link:
http://db.apache.org/derby/papers/recovery.html#Checkpoints
Hope this helps :)
---
Jørgen Løland
