i have a system with two identical drives (WDC WD800BB-00CAA0) hooked to two identical controllers (CMD649 based) in a simple mirror. randomly, the system will exhibit very high load averages, almost exclusively from blocked processes. running iostat -x at that time will show that the hde drive will have much large await and svctime values that hda. i had a different set of controllers (PDC based) in there before and saw similar problems. i used to be running ext3 on the filesystems, but have dropped back to ext2 because of theories of the journal causing periodic slowdowns. i have also tweaked bdflush via: vm.bdflush = 20 2000 128 0 60 512 60 0 0 based on some info on the web. each change i make seemed to help, but then eventually the problem comes back. so i have another theory: should i use dissimilar controllers so that the device driver code isn't being shared? i'm just wondering if the underlying code could somehow be starving the second disk while servicing the first? or am i maybe just running into plain old bandwidth issues (pci or ide)? the system is just a general isp 'everything' box. email, web, radius, mysql, etc. it is a dual 1ghz p3 with 1g of ram on some intel server board. any other great ideas? - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
