Sad things happen. Imagine a complex scenario like this:
1. Due to some issues the log backups don't work any more.
2. The filesystem /hana/log fills up.
3. Someone creative removes a log segment. Don't ask me why.
4. The HANA instance crashes.
So the HANA instance won't start again and you have a gap in the log sequence. A point-in-time recovery (PIT) is required.
1. A first PIT recovery is started with a recovery objective several hours before the contents of that deleted log segment. This was due to the fact that the HANA instance had three different timezones: GMT, the appliance time zone and the DBA's time zone. Also the PIT recovery required the flag for "initialize log area" to be set, because of that missing log segment.
2. The PIT was successful, but looking at the data reveals that the recovery time objective was wrong.
3. Now a second PIT recovery is started with the correct time objective. The HANA studio performs the recovery, but in backup.log you can see that this recovery also recovered to the exact same as the previous PIT. The "initialize log area" option from the previous PIT made me lose even more data!
Of course, this is the product of two errors (deleting a log segment and performing a PIT with the wrong time objective).
I wonder if there is a possibility to save the contents of the log segments which have not yet been backed up? Or was that second data loss inevitable?
An additional remark to SAP: It is extremely unhelpful that all HANA recoveries start with restoring the data backup to the data files. This procedure can take a long time, and all too often a recovery will fail for some unforseen reasons. So each retry is a painful exercise in patience.