> scst: ***ERROR***: New device handler's vdisk_blockio attach() failed: -2 > dev_vdisk: ***ERROR***: filp_open(/dev/disk/by-id/md-name-localhost:vm01) failed: -2 > dev_vdisk: Registering virtual vdisk_blockio device vm01 (WRITE_THROUGH, BLOCKIO) > after reboot i got a error message saying to check /var/log/boot > After the reboot I couldnt get the volume to show up to vmware for some reason So i logged into esos and saw a error message about SCST not being loaded, so i rebooted again and watched the console boot. > But it didnt look like it took effect, so i rebooted. > echo "64" >/sys/kernel/scst_tgt/targets/iscsi/iqn.xxxxxxx/QueuedCommands > one of the things i noticed was that the Command Queue was filling up at times, I was only getting about 550MB/s over the 40Gb card and Mellanox suggested a setting in the nf of COMMANDQUEUE=64 (to match vmware's iscsi queue) > at one point i was able to get the lun to present to VMware via iSER via ROCE, I was able to format the drive as vmfs and do some testing > Anyway because the system was setup for ZFS i am using the PERC H330 HBA so no hardware RAID. > I started messing with ESOS recently after really poor performance on ubuntu/zfs for vmware (24 SSD's in a dell R730) > On Wednesday, Augat 8:45:10 PM UTC-7 Gary Kendzerski wrote: When you reboot, please post your /var/log/boot file here so we can If that doesn't work, and the MD array doesn't assemble automatically 'mdadm -assemble -scan' so it reads all block devices and assemblesĪLL arrays that it finds (not just entries in the config file). I don't believe the TUI generates an '/etc/nf' file, and youĭon't need one for simple operation: The 'rc.mdraid' script does > once it saw md0 mounted and working i ran > (assume clean because i had already let the array sync and there wasnt anything on it, not even a file system) > mdadm -create -assume-clean /dev/md0 -level=10 -raid-devices=24 /dev/sd1 > I ended up going into shell and running > it did cat on /etc/nf and it was blank > When i let the array finish sync'ing it ran > after looking at the back end using the provided 'gui' it doesnt seem to be creating the MD array correctly and sync is not adding the changes to the nf > seems like this needs to be escalated to a bug for version 2.1.1 On Thu, at 6:14 PM Gary Kendzerski wrote:
0 Comments
Leave a Reply. |