Redo Log File in Oracle

The redo log file holds transaction information. Each Commit causes a flush to the log. You have multiple log files defined - one is "active" or "online", the others are "offline". Oracle rotates through these files as each gets full. Switching active files has overhead so the suggestion is that each file can hold roughly 30 minutes or transactions.

When a file gets full, it is set to Inactive and another becomes Active. The inactive file is scheduled to be archived (by a background process). If Oracle has rotated through all of the redo logs before the first has been archived, it will have to stop and wait for the archive to finish. This is a very bad situation. To overcome, you can add more log files or increase there size. (or make your transactions bigger instead of many short ones)

Since our drivers are more transactional based, they commit very often compared to SQL-only applications. I was running 100,000 record insert tests in the Btrieve driver (each Insert gets its own Commit). After a time or two of doing this, the app slowed to a crawl. Looking at the OEM (or was it the Windows Event log?), I could see warnings about redo sync.

Anyway, after I increased the redo log file size, things ran okay. This is done in the OEM: Administration -> Redo Log Groups -> Edit each one. I don't remember the original sizes but they were very small... I now have three redo logs of 2G each.