Oracle7 Server Release 7.3.4.1.0 - Production (yes, it still exists !!) on Solaris
The IO wait on this box is very high.
Top view was consistently showing an IO wait around 90% and above.
We made certain changes, like
a) Doubled redolog sizes from 10M to 20M since we found log switches occuring too frequently.
b) Increased DB_BLOCK_BUFFERS from 40,000 to 179,200 (dbblock size is 2K)
There was no respite even after above changes and IO wait remained consistently high. (It came down to 0 when we shutdown oracle to make the init changes)
On top of it, backups started failing because of
ORA-00600: internal error code, arguments: [2103], [0], [0], [1], [900], [], [], []
SMON: following errors trapped and ignored:
ORA-01575: timeout waiting for space management resource
We went into some details and made these changes:
a) Dropped and recreated Temp tablespace (since it was badly fragmented, had pctincrease 50 and different size segments).
b) Increased Sort Area Size from 1 to 4 M
c) Increased DBWRs from 1 to 6 (system has 6 CPUs, don't know why they were keeping it one)
d) Increased db latches from 3 to 18 (but later found that oracle autoadjusted it to 12 after starting, any clues why??).
e) db_file_multiblock_read_count from 8 to 32
After we made the above changes, backups are going fine and IO wait has come down a bit, but it is still abnormal at 45 to 75 %, with idle CPU hovering near to 0-4%.
Redologs are still filling at 6 to 9 minutes (although I cant find many insert/update/delete statements going on).
log_buffer 1024000
log_checkpoint_interval 10000
log_checkpoint_timeout 0
Any clues will be highly appreciated.
Thanks
Raminder