I have 8 rollback segments(4 per instance) of 6M each. I have not specifically assigned any transaction to a specific rbs. However when I query v$rollstat, I find that in 4 rbs belonging to first instance, HWM is almost equal to 50 times the optimal size. But, in the second instance, it is 26, 0.99, 0.99 and 46 times the optimal value. Optimal is 12M and is same in all 8 rbs.
Because of the number of waits, wraps, extends and shrinks, I had recently resized the rbs . While the performance has improved very much, still extends and shrinks are present in the two segments having 26 and 46 times.(and none in any of the other 6).
It appears that the other two rbs(having 0.99 HWM of optimal) are participating in transactions but very rarely. So, how can I ensure that the problem is resolved? And are there any guidelines about how many extends and shrinks can be thought as allowable?