On Thu, Sep 22, 2005 at 01:13:18PM -0400, Jan Harkes wrote:
> Actually running multiple venus processes is possible, I used that
> several times before we had realm support. We should provide at least 4
> cfs devices.

I can see usefulness of multiple Venii for cache space / rvm space usage
isolation between vservers. So, good if it actually works.

Otherwise, outside of that necessity I would rather not run multiple
cache managers.

If we are looking at Coda as a filesystem, it is one entity.
It was not designed with partial virtualization in mind, and you
most probably will end up using more resources for multiple Venii than for
a single one, due to duplicated objects in the caches.

Normally there is no need, nor support from OS to mount the same disk partition
simultaneously on several places. Coda tries hard to behave like a disk :)
May be it shouldn't, but still, using multiple Venii seems questionnable
to me.

One of my concerns is that multiple Venii imply multiple mount points
which indirectly undermines one of Coda unique strengths - globality,
which is impossible without "/coda" being kept standard.

No problem for vservers as supposedly all of them access an own /coda,
but dangerous practice otherwise.

Regards,
--
Ivan