log file initialization after db_dump/db_load

61890

Greetings,

I have the need to db_dump/db_load a large customer database in their production environment. According to the documentation on "Log File Limits", after running db_dump/db_load, I should remove all log files and restart my application. The problem I have is my application has a check to insure that at least 1 log file exists. I cannot make changes to the application in the customers environment and cannot get a release into their environment in time to fix this issue. I have experimented with "touch log.0000000001" and this appears to work. My application starts normally and after some period of activity I can see, via db_printlog, normal activity in the log file. Is this an acceptable way to "initialize" the log file?

I am using Berkeley DB version 4.1.25 running on Solaris 9.

Thanks,
Nick

522690

The code will do the right thing if the first log file is empty. That is it will write a proper log file header and go on from there. Since the system must be able to recovery even if we fail just after creating a log file, it seems like this should continue to work.

Michael Ubell
Oracle