We're started getting Short Dumps in our DEV system which, on the surface, look like Oracle errors (SQL1555="snapshot too old"). HOWEVER:
1. Oracle is logging absoutely NO errors (the alert log is clean).
2. PSAPUNDO is more than big enough.
3. UNDO_RETENTION is sufficient.
4. Redo log sizes are big enough also.
The way we've found to fix this problem is to regenerate the programs referenced in the ST22 Dumps. Thus far, we've regenerated SAPLSVIM, SAPMM06E, and SAPL0PX7. Things are okay for a day or two, and then we need to regenerate them again (only SAPLSVIM so far: we've regenerated it twice).
Everyone says they haven't made any changes that would cause this, but there was a set of transports that touched Table View Maintenance (but for a CUSTOM table). The ABAPers all say it's BASIS/Oracle, and we (BASIS/Oracle) can't find anything Oracle-related. Again, if we regenerate the program(s) everythings works fine again for a little while.
What all this looks like if you're using the system is that ANY table viewed via SM30 (and only SM30) Short Dumps with the above SQL error along with the reference that "SQL error 1555 occurred when accessing program "SAPLSVIM " part "LOAD"". Once we regenerate the program, everything works fine again.
Any ideas?? Please send them along.
Thanks,
Lara