How to reduce upgrade downtime for R12.2

• Have bigger redo logs
• Database in no archive log mode
• Disabling Auditing
• Disable flashback
• Disable TDE
• Enable Degree of Parallelism (DOP)
• Tune init.ora parameters
– SGA_MAX_SIZE
– PGA_AGGREGATE_SIZE
– db_file_multiblock_read_count
DB Considerations
• Disable custom triggers and business events
• Tablespace sizing
• Allocate parallel channels for RMAN backups
• Recycle bin = off
Application Considerations
• Use TUMS (“The Upgrade Manual Script”) to avoid tasks not relevant to the
system
• Upgrade by request
• Batch size
• AD workers
• Increase ADJREOPTS, ADJRIOPTS parameters (Ref: 557926.1)
– Avoid errors “occurred during initialization of VM, could not reserve enough space
for object heap errors”
• Distributed AD
Application Considerations
• Merge NLS patches
• Apply Help and NLS patches online
• Apply HR Global driver online
• Merge and apply product specific patches (post upgrade)
– 62 patches merged as 10 patches
– 20 hours patching time reduced to 8 hours
Automate the upgrade process as much as possible
– Use Shell scripts, adalldefaults.txt, identify the steps we can do ahead of cutover
time
• Time every activity and deal with bottlenecks
– A baseline for upgrade execution time, analyse adtimrpt.sql output
• Maintain Issues & Resolution document for quick reference
• Maintain cookbook
– Cookbook includes end to end activities along with various phases which are
sequentially listed down
Process Considerations
• Maintain Checklist
– Crosscheck the checklist before start any major tasks (like 12.1.1 or 12.1.3 upgrade
etc.) This checklist should have SGA, PGA, job_queue_processes,
parallel_max_servers, heap size etc.