Re: [SSI-users] tmpfs (Debian & SSI 1.0.0)

[John H. <john@Calva.COM>]

Aneesh Kumar wrote:
> Hi,
>
>
> On Tue, 3 Aug 2004 19:49:53 -0700, Jeff Lerman
> <jcl...@an...> wrote:
>   
>> On our three-node cluster running SSI (1.0.0) atop Debian, I see that
>> tmpfs gets mounted by every node when it boots into the cluster.  This
>> results in multiple tmpfs entries being listed when I do "df".  If a
>> cluster crashes and restarts, another entry is added.  Earlier today I
>> was up to six or seven entries.
>>     
>
> I have commited a fix to comment out the tmpfs mount from mountvirtfs script
>   
Pretty wholesale!

The basic problem is that the only cleanup of /etc/mtab is done by the
"mount -a -d #" call from rc.nodedown.  It will zap anything that looks
like it came from the node that was down, but it has no way of knowing
that a line like:

    tmpfs /dev/shm tmpfs rw 0 0
      

or even

    sysfs /sys sysfs rw 0 0
      

came from the node that went down.

Wouldn't it be nice to add a "node=#" to mtab to simplify this?  Or
would that cause horrible problems?