On Mon, Jun 14, 2004 at 03:45:56PM +0100, Cesar Garcia wrote: > ok, error... is crash again.. > > I dont understand the test result in the right way. :( > > the local deliveries are extremely quickly, about 6000 mails in less > than 1 minute when preproccesing is finish. but when preprocced and > local deliveries are at the same time then i believe that the two > proccess are not ok together. > > > i still dont know why. > > Raid controller is a MyLEX 170 with two UWSCSI 160M/s 73GB/10000rpm is > not enough I/O subsystem? :| >
Sounds like a bad raid setup. qmail uses fsync() excessively this can cause major performance issues on some file systems and may cause large seek times because of disk head movement. As your top output showed in one of your previous mails almost all processes are blocked because the are waiting for io completition. For good performance the qmail queue should be on a different disk / raid than the acctual mail store. -- :wq Claudio
