I would still rank undo & redologs together with regular data because they're crucial for survivability and consistency of information...
 
Tanel.
    • If there is a policy to place all “corporate data” on the shared storage (NAS), then surely redo, RBS/UNDO etc should not be considered “corporate data”
    • Often, the “lk<DBNAME>” mount lock file in the “$ORACLE_HOME/dbs” directory should be placed on direct-attach HDD or SAN, else ORA-27086 can happen
      • Oracle documentation mentions resolution of ORA-27806 by running unlocking commands on NetApps console, but the best solution is prevention...
  1. Not placing file-systems like user home directories, ORACLE_HOME, admin/log/trc, APPL_TOP, etc on the NAS and placing heavily-accessed datafiles on direct-attach HDD or SAN
    • Why not place low-demand file-systems on the lower-performance storage and high-demand file-systems on higher-performance storage?

Reply via email to