Excessive Buffer Waits

Folks, Wanted to check whether or not anyone else has seen this behavior on 7.10.UC1. We have a system with 6000 buffers and a 99.5%read cache and 98.5% write cache hit ratio. Onstat is showing **very*** excessive buffer waits in some conditions. When we do a : select * from big_unfragmented_takes_half_a_disk_table where primary_key = "XXX" the buffer waits begin to build up. By doing an onstat -aB repeatedly to a file, we can see that the select job is indeed waiting for a buffer, and it gives us an address for a buffer that really exists. The question is why? When we check the buffers, they are mostly clean...nobody else is on the system and the load is very low. We have the buffers cleaning beginning at 10% and continuing until 5% are dirty. Also, when this is happening, the ratio of syscpu/usercpu goes way up, to over one. This ratio usually runs about 1/4 to 1/10, indicating that the system is incurring a lot of system activity at the time. If the engine has already identified the buffer he wants to give us, and it's not dirty, why are we having to wait for it? TIA Joe -- --------------------------------------------------------------------------- Joe Lumbley(jlumbley@netcom.com) Dallas, Texas ----------------------------------------------------------------------------