• reboot the mds
• create labels and partitions, e.g. with zpool create foo device; zpool destroy foo
• sammkfs -a 64 Linux
• mkdir /test
• mount -F samfs Linux /test

monitoring throughput

#!/bin/bash


IN1=21:00:00:24:ff:37:a5:1c
IN2=21:00:00:24:ff:37:a5:1d

while [ TRUE ]
do
        t1=$( date +"%s.%N" )
        wc11=$( cat /sys/kernel/scst_tgt/targets/qla2x00t/$IN1/write_io_count_kb )
   rc11=$( cat /sys/kernel/scst_tgt/targets/qla2x00t/$IN1/read_io_count_kb )
        wc12=$( cat /sys/kernel/scst_tgt/targets/qla2x00t/$IN2/write_io_count_kb )
   rc12=$( cat /sys/kernel/scst_tgt/targets/qla2x00t/$IN2/read_io_count_kb )
        sleep 2
        t2=$( date +"%s.%N" )
        wc21=$( cat /sys/kernel/scst_tgt/targets/qla2x00t/$IN1/write_io_count_kb )
   rc21=$( cat /sys/kernel/scst_tgt/targets/qla2x00t/$IN1/read_io_count_kb )
        wc22=$( cat /sys/kernel/scst_tgt/targets/qla2x00t/$IN2/write_io_count_kb )
   rc22=$( cat /sys/kernel/scst_tgt/targets/qla2x00t/$IN2/read_io_count_kb )
        rw1=$( echo "scale=2;($wc21-$wc11)/($t2-$t1)/1000"| bc )
        rr1=$( echo "scale=2;($rc21-$rc11)/($t2-$t1)/1000"| bc )
        rw2=$( echo "scale=2;($wc22-$wc12)/($t2-$t1)/1000"| bc )
        rr2=$( echo "scale=2;($rc22-$rc12)/($t2-$t1)/1000"| bc )
       echo "write: \e[91m$rw1\e[39m Mb/s\tread: \e[92m$rr1\e[39m Mb/s write: \e[91m$rw2\e[39m Mb/s\tread: \e[92m$rr2\e[39m Mb/s"
done

Config file for multipath

HANDLER vdisk_fileio {
        DEVICE zfs {
                filename /dev/zvol/data/volume
        }
}

TARGET_DRIVER copy_manager {
        TARGET copy_manager_tgt {
                LUN 0 zfs
        }
}

TARGET_DRIVER qla2x00t {
        TARGET 21:00:00:24:ff:2b:b1:7c {
                HW_TARGET
                enabled 1
                GROUP drives {
                        LUN 0 zfs
                        INITIATOR 21:00:00:1b:32:93:d6:0f
                }
        }
        TARGET 21:00:00:24:ff:2b:b1:7d {
                HW_TARGET
                enabled 1
                GROUP drives {
                        LUN 0 zfs
                        INITIATOR 21:00:00:1b:32:92:8f:af
                }
        }
}

• scstadmin -config /etc/scst.conf shouhd now make the WWN 21:00:00:24:ff:2b:b1:7d visible on the switch.

additional zoning on the other switch (if it is on the same switch use something else but SCR04)

• alicreate "SCR04","21:00:00:24:ff:2b:b1:7d"
• zonecreate "MD3_SCR04","MD3;SCR04"
• cfgadd "SL3000_DDN_6780","MD3_SCR04"
• cfgsave
• cfgenable "SL3000_DDN_6780"

Tests

Reboot of the Linux disk array

• the FS is still accessible, probably as long as writings can be buffered and the content is in the cache
• the logs show

 offlining LUN=0x2 (trace=0x0), target=0x21a00 (trace=0x2800004)
Dec 12 08:27:01 mds-test scsi: [ID 243001 kern.info] /pci@0,0/pci10de,377@a/pci1077,15d@0/fp@0,0 (fcp1):
Dec 12 08:27:01 mds-test        offlining LUN=0x1 (trace=0x0), target=0x21a00 (trace=0x2800004)
Dec 12 08:27:01 mds-test scsi: [ID 243001 kern.info] /pci@0,0/pci10de,377@a/pci1077,15d@0/fp@0,0 (fcp1):
Dec 12 08:27:01 mds-test        offlining LUN=0x0 (trace=0x0), target=0x21a00 (trace=0x2800004)
Dec 12 08:27:01 mds-test genunix: [ID 483743 kern.info] /scsi_vhci/disk@g6237316136666137 (sd27) multipath status: failed: path 4 fp1/disk@w21000024ff26b76b,2 is offline
Dec 12 08:27:01 mds-test genunix: [ID 483743 kern.info] /scsi_vhci/disk@g3833666463373365 (sd25) multipath status: failed: path 5 fp1/disk@w21000024ff26b76b,1 is offline
Dec 12 08:27:01 mds-test genunix: [ID 483743 kern.info] /scsi_vhci/disk@g3930356635663439 (sd26) multipath status: failed: path 6 fp1/disk@w21000024ff26b76b,0 is offline
Dec 12 08:27:09 mds-test scsi: [ID 107833 kern.warning] WARNING: /scsi_vhci/disk@g3930356635663439 (sd26):
Dec 12 08:27:09 mds-test        Command failed to complete...Device is gone


Dec 12 08:29:41 mds-test fctl: [ID 517869 kern.warning] WARNING: fp(0)::N_x Port with D_ID=11a00, PWWN=21000024ff26b76a reappeared in fabric
Dec 12 08:29:42 mds-test fctl: [ID 517869 kern.warning] WARNING: fp(1)::N_x Port with D_ID=21a00, PWWN=21000024ff26b76b reappeared in fabric

exchange the QLogic card

• configure the zoning on the switch and modify the /etc/scst.conf on the target side
• the mds recognizes the new paths and lists the old ones as unavailable. Don't know yet, how to get rid of the unavailable paths.