How to speed up slow zpool scrub?

[Jeremy Faulkner]

On 2016-04-26 11:08 AM, Miroslav Lachman wrote:
> Jeremy Faulkner wrote on 04/26/2016 17:01:
>> zfs get all tank0
>
> I set checksum=fletcher4 and compression=lz4 (+ atime & exec to Off),
> anything else is in default state.
>
> There are 19 filesystems on tank0 and each have about 5 snapshots.
>
> I don't know how long scrub runs on some others system. If it is limited
> by CPU, or disk IOps... but for me 3 - 4 days are really long.
>
>
> # zfs get all tank0
> NAME   PROPERTY              VALUE                  SOURCE
> tank0  type                  filesystem             -
> tank0  creation              Thu Jul 23  1:37 2015  -
> tank0  used                  7.85T                  -
> tank0  available             2.26T                  -
> tank0  referenced            140K                   -
> tank0  compressratio         1.86x                  -
> tank0  mounted               no                     -
> tank0  quota                 none                   default
> tank0  reservation           none                   default
> tank0  recordsize            128K                   default
> tank0  mountpoint            none                   local
> tank0  sharenfs              off                    default
> tank0  checksum              fletcher4              local
> tank0  compression           lz4                    local
> tank0  atime                 off                    local
> tank0  devices               on                     default
> tank0  exec                  off                    local
> tank0  setuid                on                     default
> tank0  readonly              off                    default
> tank0  jailed                off                    default
> tank0  snapdir               hidden                 default
> tank0  aclmode               discard                default
> tank0  aclinherit            restricted             default
> tank0  canmount              on                     default
> tank0  xattr                 on                     default
> tank0  copies                1                      default
> tank0  version               5                      -
> tank0  utf8only              off                    -
> tank0  normalization         none                   -
> tank0  casesensitivity       sensitive              -
> tank0  vscan                 off                    default
> tank0  nbmand                off                    default
> tank0  sharesmb              off                    default
> tank0  refquota              none                   default
> tank0  refreservation        none                   default
> tank0  primarycache          all                    default
> tank0  secondarycache        all                    default
> tank0  usedbysnapshots       0                      -
> tank0  usedbydataset         140K                   -
> tank0  usedbychildren        7.85T                  -
> tank0  usedbyrefreservation  0                      -
> tank0  logbias               latency                default
> tank0  dedup                 off                    default
> tank0  mlslabel                                     -
> tank0  sync                  standard               default
> tank0  refcompressratio      1.00x                  -
> tank0  written               140K                   -
> tank0  logicalused           13.3T                  -
> tank0  logicalreferenced     9.50K                  -
> tank0  volmode               default                default
> tank0  filesystem_limit      none                   default
> tank0  snapshot_limit        none                   default
> tank0  filesystem_count      none                   default
> tank0  snapshot_count        none                   default
> tank0  redundant_metadata    all                    default
>

Check the drive health with smartctl (part of sysutils/smartmontools). 
Are the drives desktop drives or nas drives? In gstat is one drive 
busier than the rest? If so, replace that drive.