I have an Oracle 9.2.0.6. database running on Solaris in archivelogmode.

For some reason the UNDO tablespace is filling up and doesn't seem to empty, while the redo logfiles (after filling up) don't write to the archivelogs leading to a standstill of the instance. There is still loads of disk space that the archivelogs could be written to and the init ora parameters are set to sensible values.

After restarting the instance, the UNDO tablespace is still pretty full.

Could this have something to do with very large transactions that continue to process data beyond the capacity of the UNDO tablesoace and the redo logfiles, or what could possibly cause this?