sorry for not getting back sooner.  this seems reasonable but I just want
to make sure it's complete.  for example that when building plot files do
the headers get dealt with correctly?  and does playback work as well?  if
following the 'cciss' route, it's probably going to work and that;s
probably what I would have done too.  perhaps a switch like --diskremap
foo:bar,xxx:yyy where these are sets of names you want to remap?  in your
case you might say somethng like --diskremap etherd/: leaving off the new
name to just indicate replacing it with nothing.  make sense?
-mark


On Tue, Sep 22, 2015 at 2:29 PM, Eggleston, Gabriel (NIH/NHGRI) [C] <
[email protected]> wrote:

> Good afternoon,
>
> We're monitoring some systems with Ganglia and collectl, and are also
> using AoE. The devices are in the format "/dev/etherd/e##.#" (or
> "/dev/etherd/e##.#p#" for partitions,) and they show up in
> /proc/diskstats as "etherd/e##.#"
>
> The metrics sent by gexpr (like "diskinfo.writes.etherd/e21.0") make
> Ganglia pretty upset due to the "/" character. I see that formatit.ph
> strips "cciss/" from disk names, so I tried duplicating that with
> "etherd/" in the changes in the diff below. It's working well for us
> with this change applied - could you consider including it in the next
> version of collectl?
>
> (We currently use these devices without partitioning, but for
> consistency I tried copying the method of excluding partitions from
> cciss devices. I wasn't sure if the other cciss special-case handling in
> this file applied to the AoE devices, so I left those alone.)
>
> --
>
> --- formatit.ph 2015-09-22 10:31:21.000000000 -0400
> +++ formatit.ph.patch1  2015-09-22 11:59:30.000000000 -0400
> @@ -239,6 +239,7 @@
>      my $diskName=$fields[3];
>      $diskName=remapDiskName($diskName)    if $diskRemapFlag;
>      $diskName=~s/cciss\///;
> +    $diskName=~s/etherd\///;
>      push @dskOrder, $diskName;
>      $disks{$diskName}=$dskIndexNext++;
>    }
> @@ -697,11 +698,12 @@
>
>            # now for the magic.  Get the partition size and name, but
> ignore
>            # cciss devices on controller 0 OR any devices with partitions
> -          # noting cciss device partitions end in 'p-digit' and sd
> partitions
> +          # noting cciss/etherd device partitions end in 'p-digit' and
> sd partitions
>            # always end in a digit.
>           ($size, $name)=(split(/\s+/, $part))[3,4];
>           $name=~s/cciss\///;
> -         next    if $name=~/^c0|^c.*p\d$|^sd.*\d$/;
> +         $name=~s/etherd\///;
> +         next    if $name=~/^c0|^c.*p\d$|^e.*p\d$|^sd.*\d$/;
>            $partitionSize{$name}=$size;
>          }
>
> @@ -3278,6 +3280,7 @@
>      ($major, $minor, $diskName, @dskFields)=split(/\s+/, $data);
>
>      $diskName=~s/cciss\///;
> +    $diskName=~s/etherd\///;
>      if (!defined($disks{$diskName}))
>      {
>        $dskChangeFlag|=1;    # new disk found
> @@ -4599,6 +4602,7 @@
>
> $temp.="[DSK]QueLen${SEP}[DSK]Wait${SEP}[DSK]SvcTim${SEP}[DSK]Util${SEP}";
>        $temp=~s/DSK/DSK:$dskName/g;
>        $temp=~s/cciss\///g;
> +      $temp=~s/etherd\///g;
>        $dskHeaders.=$temp;
>      }
>      writeData(0, $ch, \$dskHeaders, DSK, $ZDSK, 'dsk', \$headersAll);
>
> --
>
> Thanks,
> Gabriel Eggleston
> NHGRI Contractor – Digicon
> [email protected]
>
>
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Collectl-interest mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/collectl-interest
>
------------------------------------------------------------------------------
_______________________________________________
Collectl-interest mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/collectl-interest

Reply via email to