Re: [Vrs-development] IBM Token Ring Thingy...

[Chris Smith]

On Saturday 04 May 2002 00:45, Bill Lance wrote:
> --- address@hidden wrote:
> > I agree with the looseness of LDS within a cluster
> > but a 2 phase commit will
> > provide better reliability and rollback support.
>
> How can we do it without locking up processes?

Well, you're running within a middleware.
The messages between the component parts of your LDS and VRS are middleware 
messages.  These messages are subject to timeouts, so there is no way that 
you can lock up a process indefinately.

Once you start a transaction and then access the Resource Manager at LDS(B), 
(depending on your granuality of locking) the RM will be locked until the 
transaction is committed or rolled back.

If a timeout occurs, the middleware will automatically rollback the 
transaction that the timed-out request was part.

RM Requests which rely on a lock will queue in a FIFO whilst existing locks 
are satisfied.

If a request is in the FIFO for too long, then the request will timeout, and 
cause the transaction that it was part of to rollback.

See!  I told you using a Middleware was a 'Good Idea' (tm).

You're going to want me to put transaction support in aren't you?
Gits!

It's partly there now (well I've left some convenient hooks in place)....

I'll have a think...

Chris

--
Chris Smith
  Technical Architect - netFluid Technology Limited.
  "Internet Technologies, Distributed Systems and Tuxedo Consultancy"
  E: address@hidden  W: http://www.nfluid.co.uk