After a migration of a database to a faster machine the database was slow and appears to hung. The problem was a slow and big lob segment. The following bug was the reason for the behaviour:
Bug 6376915: ENQ: HW - CONTENTION WITH LOB SEGMENTS.
The bigger the lob segment, the more concurrent CPUs, the more write and read I/O => the bigger the problem.
Bug 6376915: ENQ: HW - CONTENTION WITH LOB SEGMENTS.
The bigger the lob segment, the more concurrent CPUs, the more write and read I/O => the bigger the problem.
Oracle LOBs + ASSM + many CPUs = enq: HW – contention
Our new hardware architecture had significant more cores anf threads per CPU. The problem already existed with the old hardware but the new hardware made it really worse. Take a look at the statspack report:
Top 5 Timed Events Avg
%Total
~~~~~~~~~~~~~~~~~~
wait Call
Event Waits Time (s)
(ms) Time Wait Class
------------------------------
------------ ----------- ------ ------ ----------
enq: HW - contention 224,968 658,323
2926 93.1 Configurat
