1. The Log is used to store information that's used to rollback or replay transactions
2. The Log grows until the data is commited and ?safely? written into the data-tables
That is not correct.
Because during commit not every data page needed to store the data
changed by that transaction is written to disk and therefore save,
the log is NOT cleared when the commit finishes.
It is needed for those cases where between the commit and the next
time EVERY changed data page is written to disk, the kernel is stopped
(without shutdown, which would write to disk).
I took a look again at the AutoLog Stuff in the DatabaseManager.
I have to setup a mediuma, a file for example.
If SAPDB feels like it should a Log-backup, it writes the current Log to the file. But what if the file already exists? Will it be overwritten?
I marked the "overwrite"-checkbox of the backup-file, but what if this is disabled?
will the autolog-thing work only once than?
Or does the AutoLog feature append the new backup to the file?

To assure after restart that every committed transaction can be redone
one has to store the log even after commit and save it using explicit or
implicit (auto-) backup-mechanism.
And how does the Kernel access the backuped logs? It just expects, that the old file are there to restore? what if i move them away? what would i have to do then?

Will the kernel ever delete old autolog-backups it doesn't need anymore?
How can i decide which Log-Backups aren't needed anymore?



_______________________________________________
sapdb.general mailing list
[EMAIL PROTECTED]
http://listserv.sap.com/mailman/listinfo/sapdb.general


Reply via email to