Mailing List Archive

On 2000-02-28T06:30:19,
Marcelo Tosatti <marcelo@conectiva.com.br> said:

> cluster manager is needed. So it would be interesting to create a generic
> API which DRBD 2.0 (and other cluster applications) could use to be
> independant of the underlying cluster services API. But first, we have to
> analyze what parts of it are possible.

If that would be possible, it would be great.

However, I fear the semantics of different cluster cores may be too different,
or you lose access to the "advanced" features they offer.

Which cores are you looking at? Ensemble I gather, the LCC solution. If you
are at it, please, check out
http://www.sgi.com/software/failsafe/tech_info.html too, and give me some
feedback if you think it could be of any help to your product, or why it would
be no help at all.

(FailSafe will probably appear on Linux very soon)

Sincerely,
Lars Marowsky-Brée

--

Hi,

On Mon, 28 Feb 2000 06:30:19 -0300 (BRST), Marcelo Tosatti <marcelo@conectiva.com.br> said:

> To start DRBD 2.0 development (Jerome will send a draft of it soon to
> the list) a cluster manager will be needed. Since Stephen's cluster
> manager (BTW, it has a name already? :)) will not be ready so soon,
> another cluster manager is needed. So it would be interesting to
> create a generic API which DRBD 2.0 (and other cluster applications)
> could use to be independant of the underlying cluster services
> API. But first, we have to analyze what parts of it are
> possible. Stephen, could you please send me your cluster API
> definition and CC it to the list?

The bits I've got written up were posted by Mike Wangsmo a few weeks
ago, to ftp://people.redhat.com/wanger/phase_II. Right now, everybody
is holding their breath waiting to see what SGI come up with in
FailSafe.

--Stephen