#!/bin/bash echo "$0 - 2021 Dave Bechtel - make a ZFS DRAID pool" echo "- pass arg1='reset' to destroy test pool" echo "- pass arg1='fail' and arg2=dev2fail to simulate failure" echo "Reboot to clear simulated device failures before issuing 'reset'" # Requires at least zfs 2.1.0 DD=/dev/disk DBI=/dev/disk/by-id # total disks for pool / children td=24 # 26 - root and 1 spare # raidz level (usually 2) rzl=1 # spares - per vdev spr=1 # TODO EDITME zp=zdraidtest function zps () { zpool status -v |awk 'NF>0' } #pooldisks=$(echo /dev/sd{b..y}) #pooldisks1=$(echo /dev/sd{b..m}) #pooldisks2=$(echo /dev/sd{n..y}) #pooldisks=$pooldisks1' '$pooldisks2 # need entire set for reset # 24 disks = groups of 8 pooldisks1=$(echo /dev/sd{b..i}) # bcdefghi pooldisks2=$(echo /dev/sd{j..q}) # jklmnopq pooldisks3=$(echo /dev/sd{r..y}) # rstuvwxy # z is reserved phys spare pooldisks=$pooldisks1' '$pooldisks2' '$pooldisks3 # need entire set for reset #pooldisks=$pooldisks1' '$pooldisks2' '$pooldisks3' '$pooldisks4' '$pooldisks5' '$pooldisks6 # need entire set for reset # 24, groups of 6 drives = 4 vdevs # 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 # sdb sdc sdd sde sdf sdg sdh sdi sdj sdk sdl sdm sdn sdo sdp sdq sdr sds sdt sdu sdv sdw sdx sdy # D D D Z2 Z2 S = draid2:3d:6'c':1's' # D D D D Z1 S = draid1:4d:6'c':1's' # D D D D D Z1 (no vspare) = draid1:5d:6'c':0's' # extending to 32 disks #pooldisks2=$(echo /dev/sda{a..h}) #sdaa sdab sdac sdad sdae sdaf sdag sdah # failexit.mrg function failexit () { echo '! Something failed! Code: '"$1 $2" # code # (and optional description) exit $1 } # cre8 drive translation table - NOTE 32 disk config gets overridden vv source ~/bin/boojum/draid-pooldisks-assoc.sh $td # Flame the pool and start over from 0 if [ "$1" = "reset" ]; then # no need to worry if its not imported / already destroyed if [ $(zpool list |grep -c $zp) -gt 0 ]; then zpool destroy $zp || failexit 999 "Failed to destroy $zp" fi for d in $pooldisks; do echo -e -n "o Clearing label for disk $d \r" zpool labelclear -f "$d"1 done echo '' # also reset hotspares # echo ${hotspares[@]} # zpool status -v |egrep 'sdz|sday|sdaz|sdby|sdbz|sdcy|sdcz' for d in ${hotspares[@]}; do #echo $d # DEBUG echo -e -n "o Clearing label for Hotspare disk $d \r" zpool labelclear -f "/dev/$d"1 done echo '' zpool status -v exit; # early fi # Simulate a drive failure; if zed daemon is running, a spare should auto kick in if [ "$1" = "fail" ]; then # NOTE we do NO error checking here, so if you fail your ROOT DISK, THAT'S ON YOU! echo "$(date) - $0 - Simulating disk failure for $2 $(ls -lR $DD |grep $2)" |tee |logger echo offline > /sys/block/$2/device/state cat /sys/block/$2/device/state |tee |logger time dd if=/dev/urandom of=/$zp/^^tmpfileDELME bs=1M count=$td; sync # force a write; if not work, try scrub zps exit; # early fi # zpool create draid[][:d][:c][:s] # ex: draid2:4d:1s:11c # data - The number of data devices per redundancy group # In general a smaller value of D will increase IOPS, improve the compression # ratio, and speed up resilvering at the expense of total usable capacity. # SLOW writing to zstd-3 # draid$rzl:8d:12'c':$spr's' $pooldisks1 \ # draid$rzl:8d:12'c':$spr's' $pooldisks2 \ # handy REF: https://arstechnica.com/gadgets/2021/07/a-deep-dive-into-openzfs-2-1s-new-distributed-raid-topology/ # groups of 8 drives = 3 vdevs (z is reserved for physical hotspare) # sdb sdc sdd sde sdf sdg sdh sdi sdj sdk sdl sdm sdn sdo sdp sdq sdr sds sdt sdu sdv sdw sdx sdy # 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 # D D D D D Z2 Z2 S = draid2:5d:8'c':1's' # D D D D D D Z1 S = draid1:6d:8'c':1's' # D D D D D D D Z1 (no vspare) = draid1:7d:8'c':0's' # TODO EDITME iteration=1 if [ "$iteration" = "1" ]; then # compression=zstd-3 # -o ashift=12 # NOTE will NOT do 7d ( set -x time zpool create -o autoexpand=on -o autoreplace=on -O atime=off -O compression=lz4 \ $zp \ draid$rzl:6d:8'c':$spr's' $pooldisks1 \ draid$rzl:6d:8'c':$spr's' $pooldisks2 \ draid$rzl:6d:8'c':$spr's' $pooldisks3 \ || failexit 101 "Failed to create DRAID" ) elif [ "$iteration" = "2" ]; then # raidz level (usually 2) rzl=2 # spares - per vdev spr=1 ( set -x time zpool create -o autoexpand=on -o autoreplace=on -O atime=off -O compression=lz4 \ $zp \ draid$rzl:5d:8'c':$spr's' $pooldisks1 \ draid$rzl:5d:8'c':$spr's' $pooldisks2 \ draid$rzl:5d:8'c':$spr's' $pooldisks3 \ || failexit 101 "Failed to create DRAID" ) elif [ "$iteration" = "3" ]; then # This appears to be a "useless" config, you gain nothing apparent by going down to 5D # raidz level (usually 2) rzl=1 # spares - per vdev spr=1 ( set -x time zpool create -o autoexpand=on -o autoreplace=on -O atime=off -O compression=lz4 \ $zp \ draid$rzl:5d:8'c':$spr's' $pooldisks1 \ draid$rzl:5d:8'c':$spr's' $pooldisks2 \ draid$rzl:5d:8'c':$spr's' $pooldisks3 \ || failexit 101 "Failed to create DRAID" ) else # One Big Mother, 1 vspare, 0x pspares # -o ashift=12 # raidz level (usually 2) rzl=1 # spares spr=1 ( set -x time zpool create -o autoexpand=on -O atime=off -O compression=lz4 \ $zp \ draid$rzl:8d:$td'c':$spr's' $pooldisks \ || failexit 101 "Failed to create DRAID" ) #rc=$? #[ $rc -gt 0 ] && exit $rc fi rc=$? [ $rc -gt 0 ] && exit $rc # ^ Need this check because of subshell, will not exit early otherwise #[ $(zpool list |grep -c "no pools") -eq 0 ] && \ # zpool add $zp spare ${hotspares[@]} # The below will not work: gets error # "requested number of dRAID data disks per group 6 is too high, at most 3 disks are available for data" #( set -x #time zpool create -o ashift=12 -o autoexpand=on -O atime=off -O compression=lz4 \ # $zp \ # draid$rzl:6d:6'c':$spr's' $pooldisks1 \ # draid$rzl:6d:6'c':$spr's' $pooldisks2 \ # draid$rzl:6d:6'c':$spr's' $pooldisks3 \ # draid$rzl:6d:6'c':$spr's' $pooldisks4 \ #|| failexit 101 "Failed to create DRAID" #) # requires external script in the same PATH # going with lz4 so not limited by CPU for compression zfs-newds.sh 11 $zp shrcompr zfs-newds.sh 10 $zp notshrcompr zfs-newds-zstd.sh 10 $zp notshrcompr-zstd zfs-newds.sh 00 $zp notshrnotcompr zps zpool list zfs list df -hT |egrep 'ilesystem|zfs' echo "NOTE - best practice is to export the pool and # zpool import -a -d $DBI" date exit; # REFS: https://openzfs.github.io/openzfs-docs/Basic%20Concepts/dRAID%20Howto.html https://klarasystems.com/articles/openzfs-draid-finally/ https://www.reddit.com/r/zfs/comments/lnoh7v/im_trying_to_understand_how_draid_works_but_im/ https://insider-voice.com/a-deep-dive-into-the-new-openzfs-2-1-distributed-raid-topology/ https://docs.google.com/presentation/d/1uo0nBfY84HIhEqGWEx-Tbm8fPbJKtIP3ICo4toOPcJo/edit#slide=id.g9d6b9fd59f_0_27 Group size must divide evenly into draid size E.g., 30 drives can only support 3 drive group 5 drive group 10 drive group 15 drive group Only need to specify group size at creation Group Size - the number of pieces the data is partitioned into plus the amount of parity o The amount of parity determines the redundancy o The number of data pieces determines the overhead dRAID Size - the number of drives used for data (Does not include spare drives) ----- # make a OneBigMother 24-disk draid with raidz1, 1 VDEV, 8 data disks, 24 children, 0 vspare + 5 pspares + zpool create -o autoexpand=on -O atime=off -O compression=lz4 zdraidtest \ draid1:8d:24c:0s \ /dev/sdb /dev/sdc /dev/sdd /dev/sde /dev/sdf /dev/sdg /dev/sdh /dev/sdi \ /dev/sdj /dev/sdk /dev/sdl /dev/sdm /dev/sdn /dev/sdo /dev/sdp /dev/sdq \ /dev/sdr /dev/sds /dev/sdt /dev/sdu /dev/sdv /dev/sdw /dev/sdx /dev/sdy real 0m3.210s Dumping shortdisk == longdisk assoc array to /tmp/draid-pooldisks-assoc.log + zfs create -o atime=off -o compression=lz4 -o sharesmb=on -o xattr=sa -o recordsize=1024k zdraidtest/shrcompr cannot share 'zdraidtest/shrcompr: system error': SMB share creation failed filesystem successfully created, but not shared changed ownership of '/zdraidtest/shrcompr' from root to user Filesystem Type Size Used Avail Use% Mounted on zdraidtest/shrcompr zfs 76G 1.0M 76G 1% /zdraidtest/shrcompr + zfs create -o atime=off -o compression=lz4 -o sharesmb=off -o recordsize=1024k zdraidtest/notshrcompr changed ownership of '/zdraidtest/notshrcompr' from root to user Filesystem Type Size Used Avail Use% Mounted on zdraidtest/notshrcompr zfs 76G 1.0M 76G 1% /zdraidtest/notshrcompr + zfs create -o atime=off -o compression=off -o sharesmb=off -o recordsize=1024k zdraidtest/notshrnotcompr changed ownership of '/zdraidtest/notshrnotcompr' from root to user Filesystem Type Size Used Avail Use% Mounted on zdraidtest/notshrnotcompr zfs 76G 1.0M 76G 1% /zdraidtest/notshrnotcompr pool: zdraidtest state: ONLINE config: NAME STATE READ WRITE CKSUM zdraidtest ONLINE 0 0 0 draid1:8d:24c:0s-0 ONLINE 0 0 0 sdb ONLINE 0 0 0 sdc ONLINE 0 0 0 sdd ONLINE 0 0 0 sde ONLINE 0 0 0 sdf ONLINE 0 0 0 sdg ONLINE 0 0 0 sdh ONLINE 0 0 0 sdi ONLINE 0 0 0 sdj ONLINE 0 0 0 sdk ONLINE 0 0 0 sdl ONLINE 0 0 0 sdm ONLINE 0 0 0 sdn ONLINE 0 0 0 sdo ONLINE 0 0 0 sdp ONLINE 0 0 0 sdq ONLINE 0 0 0 sdr ONLINE 0 0 0 sds ONLINE 0 0 0 sdt ONLINE 0 0 0 sdu ONLINE 0 0 0 sdv ONLINE 0 0 0 sdw ONLINE 0 0 0 sdx ONLINE 0 0 0 sdy ONLINE 0 0 0 spares sdz AVAIL sday AVAIL sdaz AVAIL sdby AVAIL sdbz AVAIL errors: No known data errors NAME SIZE ALLOC FREE CKPOINT EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT zdraidtest 88.0G 1.52M 88.0G - - 0% 0% 1.00x ONLINE - NAME USED AVAIL REFER MOUNTPOINT zdraidtest 948K 75.8G 96.0K /zdraidtest zdraidtest/notshrcompr 96.0K 75.8G 96.0K /zdraidtest/notshrcompr zdraidtest/notshrnotcompr 96.0K 75.8G 96.0K /zdraidtest/notshrnotcompr zdraidtest/shrcompr 96.0K 75.8G 96.0K /zdraidtest/shrcompr Filesystem Type Size Used Avail Use% Mounted on zdraidtest zfs 76G 128K 76G 1% /zdraidtest zdraidtest/shrcompr zfs 76G 1.0M 76G 1% /zdraidtest/shrcompr zdraidtest/notshrcompr zfs 76G 1.0M 76G 1% /zdraidtest/notshrcompr zdraidtest/notshrnotcompr zfs 76G 1.0M 76G 1% /zdraidtest/notshrnotcompr NOTE - best practice is to export the pool and # zpool import -a -d /dev/disk/by-id Spares for this configuration can be used for the entire pool. ----- Below is a severely degraded pool with all physical spares in use; despite the raidz1 level it has sustained (6) simultaneous drive failures; one more with no replacements will kill the pool and there are no vspares allocated, but it was configured for maximum available space: pool: zdraidtest state: DEGRADED status: One or more devices could not be used because the label is missing or invalid. Sufficient replicas exist for the pool to continue functioning in a degraded state. action: Replace the device using 'zpool replace'. see: https://openzfs.github.io/openzfs-docs/msg/ZFS-8000-4J scan: scrub repaired 0B in 00:00:05 with 0 errors on Mon Jul 5 12:31:42 2021 config: NAME STATE READ WRITE CKSUM zdraidtest DEGRADED 0 0 0 draid1:8d:24c:0s-0 DEGRADED 0 0 0 spare-0 DEGRADED 0 0 0 sdb UNAVAIL 0 0 0 sdz ONLINE 0 0 0 sdc ONLINE 0 0 0 spare-2 DEGRADED 0 0 0 sdd UNAVAIL 0 0 0 sday ONLINE 0 0 0 sde ONLINE 0 0 0 spare-4 DEGRADED 0 0 0 sdf UNAVAIL 0 0 0 sdaz ONLINE 0 0 0 sdg ONLINE 0 0 0 spare-6 DEGRADED 0 0 0 sdh UNAVAIL 0 0 0 sdby ONLINE 0 0 0 sdi ONLINE 0 0 0 spare-8 DEGRADED 0 0 0 sdj UNAVAIL 0 0 0 sdbz ONLINE 0 0 0 sdk ONLINE 0 0 0 sdl ONLINE 0 0 0 sdm ONLINE 0 0 0 sdn ONLINE 0 0 0 sdo ONLINE 0 0 0 sdp ONLINE 0 0 0 sdq ONLINE 0 0 0 sdr ONLINE 0 0 0 sds ONLINE 0 0 0 sdt ONLINE 0 0 0 sdu ONLINE 0 0 0 sdv ONLINE 0 0 0 sdw ONLINE 0 0 0 sdx ONLINE 0 0 0 sdy UNAVAIL 0 0 0 spares sdz INUSE currently in use sday INUSE currently in use sdaz INUSE currently in use sdby INUSE currently in use sdbz INUSE currently in use errors: No known data errors ----- NOTE if you simulate/take a drive offline, you cant just "echo online" to it later, that wont bring it back up! try rescan-scsi-bus.sh or reboot FIX: if a drive is offline, replace it temporarily with a builtin spare: # zpool replace zdraidtest sdd draid2-0-0 # zps pool: zdraidtest state: DEGRADED status: One or more devices could not be used because the label is missing or invalid. Sufficient replicas exist for the pool to continue functioning in a degraded state. action: Replace the device using 'zpool replace'. see: https://openzfs.github.io/openzfs-docs/msg/ZFS-8000-4J scan: resilvered 0B in 00:00:00 with 0 errors on Sat Jul 3 14:43:51 2021 config: NAME STATE READ WRITE CKSUM zdraidtest DEGRADED 0 0 0 draid2:5d:24c:2s-0 DEGRADED 0 0 0 sdb ONLINE 0 0 0 sdc ONLINE 0 0 0 spare-2 DEGRADED 0 0 0 sdd UNAVAIL 0 0 0 draid2-0-0 ONLINE 0 0 0 sde ONLINE 0 0 0 sdf ONLINE 0 0 0 sdg ONLINE 0 0 0 sdh ONLINE 0 0 0 sdi ONLINE 0 0 0 sdj ONLINE 0 0 0 sdk ONLINE 0 0 0 sdl ONLINE 0 0 0 sdm ONLINE 0 0 0 sdn ONLINE 0 0 0 sdo ONLINE 0 0 0 sdp ONLINE 0 0 0 sdq ONLINE 0 0 0 sdr ONLINE 0 0 0 sds ONLINE 0 0 0 sdt ONLINE 0 0 0 sdu ONLINE 0 0 0 sdv ONLINE 0 0 0 sdw ONLINE 0 0 0 sdx ONLINE 0 0 0 sdy ONLINE 0 0 0 spares draid2-0-0 INUSE currently in use draid2-0-1 AVAIL errors: No known data errors HOWTO fix the above situation with the same disk (you rebooted / it came back online) and decouple the in-use spare: zpool export -a fdisk -l /dev/sdd # scsi-SATA_VBOX_HARDDISK_VBbcc6c97e-f68b8368 zpool labelclear /dev/sdd zpool labelclear -f /dev/sdd1 zpool import -a zpool status -v # This will show a degraded pool with a missing disk # This wont work but gives useful info: zpool replace zdraidtest spare-2 scsi-SATA_VBOX_HARDDISK_VBbcc6c97e-f68b8368 # got error, use detach zpool detach zdraidtest 2582498653363374334 # this was listed as UNAVAIL with the spare in-use underneath it zpool status -v # should now show only the spare where sdd was # we labelcleared it so it should be ready for re-use; # if you want to be really thorough you can DD zeros to the entire drive but not really necessary zpool replace zdraidtest draid2-0-0 scsi-SATA_VBOX_HARDDISK_VBbcc6c97e-f68b8368 # same disk (sdd) but labelcleared zpool status -v