Make sure the kernel DM event model works (with "dmsetup wait") If not, revert the Mike Christie patches from the udm patchset. regards, cvaroqui Selon Nicola Ranaldo <ranaldo@xxxxxxxx>: > Hello... > I have an hp proliant with qlocic 2312 connected on a dual hs80 storage > controller in multibus failover configuration linux kernel 2.6.9 with the > latest dm patch applied. > I modified kernel scsi_devinfo.c and deleted BLIST_NOSTARTONADD in order to > get capacity for ghost paths. > So when i do a modprobe qla2300 i get: > /dev/sda spinned 36gb > /dev/sdd initially ghost spinned 36gb > > so after i have > > /dev/sda ghost > /dev/sdd ok > > i do dmsetup create volume1 table > where table contains: > "0 71114623 multipath 2 round-robin 1 0 /dev/sdd round-robin 1 0 /dev/sda" > mkreiserfs /dev/mapper/volume1 > mount /dev/mapper/volume1 /mnt > cp /dev/urandom /mnt/foo > > now: > if go on the hsg80 controller serial console and give "restart other" in > order to restart the controller with active path i get: > REISER-FS: panic (device-dm-0): journal-601, buffer write failed > > another example: > if i do an sg_start /dev/sda, sdd dies but the "cp" continues to work, but > after this partial working failover, if i do an sg_start /dev/sdd i get > panic again... > > wath do i miss? > (sorry for english) > > thanks > > Nicola Ranaldo > > > > -- > > dm-devel@xxxxxxxxxx > https://www.redhat.com/mailman/listinfo/dm-devel > --