[Cluster-devel] [PATCH 1/2] fence_scsi: move key file to /var/run/cluster

[Fabio M. Di Nitto <fabbione@fabbione.net>]

Only one comment here, before attempting to read or write to
/var/run/cluster, your script needs to make sure that the directory
exists (same for the other patch).

/var/run/cluster is created only via cman init at the moment and if
fence_scsi is used outside that context, it will fail on distribution
where /var/run is on tmpfs and cluster/ does not exists.

some distributions are deploying methods to populate
/var/run/some_dir_name at boot time, but they are not consistent enough
to be trusted to behave all the same way yet.

Fabio

On 3/1/2011 4:55 PM, Ryan O'Hara wrote:
> Since the contents of /var/run/cluster/ directory are cleared on each
> reboot, move the fence_scsi.key file to that directory. This is useful
> for watchdog checks as a means to avoid watchdog reboots when a node has
> not yet unfenced or unfencing fails.
> 
> Signed-off-by: Ryan O'Hara <rohara@redhat.com>
> ---
>  fence/agents/scsi/fence_scsi.pl |    4 ++--
>  1 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/fence/agents/scsi/fence_scsi.pl b/fence/agents/scsi/fence_scsi.pl
> index ae658ab..e162dfe 100644
> --- a/fence/agents/scsi/fence_scsi.pl
> +++ b/fence/agents/scsi/fence_scsi.pl
> @@ -342,7 +342,7 @@ sub key_read ()
>      my $self = (caller(0))[3];
>      my $key;
>  
> -    open (\*FILE, "</var/lib/cluster/fence_scsi.key") or die "$!\n";
> +    open (\*FILE, "</var/run/cluster/fence_scsi.key") or die "$!\n";
>      chomp ($key = <FILE>);
>      close (FILE);
>  
> @@ -353,7 +353,7 @@ sub key_write ($)
>  {
>      my $self = (caller(0))[3];
>  
> -    open (\*FILE, ">/var/lib/cluster/fence_scsi.key") or die "$!\n";
> +    open (\*FILE, ">/var/run/cluster/fence_scsi.key") or die "$!\n";
>      print FILE "$_[0]\n";
>      close (FILE);
>