zabbix-agent-addons/zabbix_scripts/check_zfs

#!/usr/bin/perl -w

use strict;
use warnings;
use JSON;
use File::Which;
use Getopt::Long;

my $json       = {};
my $pool       = undef;
my $dataset    = undef;
my $sanoidmon  = undef;
my $stats      = undef;
my $pretty     = 0;

GetOptions(
  "zpool|pool=s" => \$pool,
  "dataset=s"    => \$dataset,
  "sanoid=s"     => \$sanoidmon,
  "stats=s"      => \$stats,
  "pretty"       => \$pretty
);

my $zpool  = which('zpool');
my $zfs    = which('zfs');
my $sanoid = which('sanoid');

if (not $zpool or not $zfs){
  print 'ZBX_NOTSUPPOTED';
  exit 0;
}
if (defined $sanoidmon and not $sanoid){
  die 'ZBX_NOTSUPPOTED';
}
if (defined $sanoidmon and not grep { $_ eq $sanoidmon } qw(snapshot capacity health)){
  die 'ZBX_NOTSUPPOTED';
}

if (not $pool and not $dataset and not $sanoidmon and not $stats){
  print <<_EOF;
Usage:
$0 [--zpool=<name>|--dataset=<fs zvol or snap>|--sanoid=<snapshot|capacity|health>]
_EOF
  exit 1;
}

# Value map. For Zabbix, we want 0 instead of none
# We also prefer on/off represented as 1/0 as it's more efficient
my $map = {
  18446744073709551615 => 0, # See https://github.com/zfsonlinux/zfs/issues/9306
  none                 => 0,
  on                   => 1,
  off                  => 0
};

if ($pool){
  foreach (qx($zpool get all $pool -p -H)){
    chomp;
    my @parse = split /\t+/, $_;
    $json->{$parse[1]} = (defined $map->{$parse[2]}) ? $map->{$parse[2]} : $parse[2];
    $json->{errors} = get_zpool_errors($pool);
    $json->{stats}  = get_zpool_stats($pool);
  }
} elsif ($dataset){
  # Convert %40 back to @ (we send them as %40 in the discovery because @ is not allowed in item keys
  $dataset =~ s/%40/\@/g;
  foreach (qx($zfs get all $dataset -p -H)){
    chomp;
    my @parse = split /\t+/, $_;
    $json->{$parse[1]} = (defined $map->{$parse[2]}) ? $map->{$parse[2]} : $parse[2];
    if ($parse[1] =~ m/compressratio$/){
      # Remove trailing x for compressratio and refcompressratio as before 0.8.0 it can be like 1.23x
      $json->{$parse[1]} =~ s/x$//;
    }
  }
} elsif ($sanoidmon){
  print qx($sanoid --monitor-$sanoidmon);
  exit $?;
} elsif ($stats){
  if (not -e '/proc/spl/kstat/zfs/' . $stats){
    print 'ZBX_NOTSUPPORTED';
    exit 0;
  }
  open STATS, '</proc/spl/kstat/zfs/' . $stats;
  while (<STATS>){
    next unless (m/^(\w+)\s+4\s+(\d+)$/);
    $json->{$1} = $2;
  }
}

print to_json($json, { pretty => $pretty }) . "\n";
exit 0;

sub get_zpool_errors {
  my $pool = shift;
  my $errors = {
    read_errors  => 0,
    write_errors => 0,
    cksum_errors => 0
  };
  my $i = 0;
  my $index = {};
  foreach my $line (qx($zpool status $pool 2>/dev/null)){
    # Output looks like
    #   pool: rpool
    #  state: ONLINE
    # status: One or more devices has experienced an unrecoverable error.  An
    #         attempt was made to correct the error.  Applications are unaffected.
    # action: Determine if the device needs to be replaced, and clear the errors
    #         using 'zpool clear' or replace the device with 'zpool replace'.
    #    see: http://zfsonlinux.org/msg/ZFS-8000-9P
    #   scan: scrub repaired 0B in 0h5m with 0 errors on Tue May 29 10:04:31 2018
    # config:
    #
    #         NAME        STATE     READ WRITE CKSUM
    #          rpool       ONLINE       0     0     0
    #            mirror-0  ONLINE       0     0     0
    #              sda2    ONLINE       0     0     0
    #              sdb2    ONLINE       0     0   474
    #
    # errors: No known data errors

    # We want to save status, action, scan and errors
    if ($line =~ m/^\s*(scan|action|status|errors):\s+(\w+.*)/){
      $errors->{$1} = $2;
      $index->{$i} = $1;
    } elsif ($line !~ /:/ and defined $index->{$i-1}){
      # Here, we reconstitute multiline values (like status and action)
      chomp($line);
      $line =~ s/\s+/ /g;
      $errors->{$index->{$i-1}} .= $line;
    } elsif ($line =~ m/\s+[a-zA-Z0-9_\-]+\s+[A-Z]+\s+(?<read>\d+(\.\d+)?)(?<read_suffix>[KMT])?\s+(?<write>\d+(\.\d+)?)(?<write_suffix>[KMT])?\s+(?<cksum>\d+(\.\d+)?)(?<cksum_suffix>[KMT])?/){
      # And here, we count the number of read, write and checksum errors
      # Note that on ZoL 0.8.0 we could use zpool status -p to get rid of the suffixes
      # But -p is not supported on 0.7 and earlier, so, we just convert them manually
      $errors->{read_errors}  += convert_suffix($+{'read'},$+{'read_suffix'});
      $errors->{write_errors} += convert_suffix($+{'write'},$+{'write_suffix'});
      $errors->{cksum_errors} += convert_suffix($+{'write'},$+{'write_suffix'});
    }
    $i++;
  }
  # Ensure evey item returns something
  $errors->{$_} ||= '' foreach (qw(scan action status errors));
  return $errors;
}

# Error counter can be suffixed. Apply this suffix to get raw error numbers
sub convert_suffix {
  my $val = shift;
  my $suf = shift;
  if (!$suf){
    return $val;
  } elsif ($suf eq 'K'){
    $val *= 1000;
  } elsif ($suf eq 'M') {
    $val *= 1000000;
  } elsif ($suf eq 'T') {
    $val *= 1000000000;
  }
  return $val;
}

sub get_zpool_stats {
  my $pool = shift;
  my $stats = {};
  open UPTIME, "</proc/uptime";
  $_ = <UPTIME>;
  chomp;
  my ($uptime , undef) = split;
  $uptime = int $uptime;
  close UPTIME;
  foreach my $line (qx($zpool iostat $pool -pH)){
    if ($line =~ m/^$pool\s+\d+\s+\d+\s+(?<reads>\d+)\s+(?<writes>\d+)\s+(?<nread>\d+)\s+(?<nwritten>\d+)/){
      # zpool iostat shows average IO since boot, so just multiply it
      # by the uptime in seconds to get cumulated IO since boot
      # Zabbix server will then be able to calculate the delta between two values
      $stats->{$_} = $+{$_} * $uptime foreach (keys %+);
      last;
    }
  }
  return $stats;
}