[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Summary [was: RE: Starting a discussion on graceful shutdown]



Hi Adrian,
 
This time it's crystal clear on what you were referring to in your comments. Please see my reply in-line.
 
Thanks
 
Regards... Zafar

From: Adrian Farrel [mailto:adrian@olddog.co.uk]
Sent: Monday, February 21, 2005 7:19 AM
To: Zafar Ali; ccamp@ops.ietf.org
Subject: Re: Summary [was: RE: Starting a discussion on graceful shutdown]

Hi,
 
Thanks for answers. Snipped down to just one point.
 
> > > > Does it matter whether it is the control plane or the data 
> > > > plane that will be taken out of service?
> > >
> > > Yes, if data plane is going OOS, we are loosing traffic and
> > > motivation behind mb4b to reduce traffic hit is lost.
> >
> > Let me rephrase the question.
> > If I know that the control plane is going OOS and I want to
> > continue to be able to manage my LSPs I will want to move
> > them to LSRs that will continue to have an active control
> > plane. The data plane through the impacted LSRs may continue
> > to be active (and therefore could continue to carry traffic).
>
> I am not sure if I understood your comment completely. However,
> the ID mentions about Grace Period and Removal of data plane
> associated with the LSP under mb4b.
 
I'll try again.
The draft talks about cases where the data plane is going to be removed. No problem there.
I am asking about cases where we know that the control plane is going to be removed, but where we know that the data plane is going to stay up.
Examples include:
- going to take the software components down (perhaps for upgrade)
- going to remove the control plane CPU (to replace it, dust it, or fit nerw tires)
- going to take down the control channel (e.g. decommission the OSC laser) 
 
Yes, most certainly, this is one of the motivations behind this draft. In fact the -00 version of the ID had following statement in the Introduction section.
 
"a Service Provider may desire to gracefully (temporarily or definitely) remove a TE Link, a group of TE Links or an entire node for administrative reasons such as link maintenance or LSR software/hardware upgrade. "
 
The revised version of the ID (-01), already in draft queue have this as one of the main requirements.
 
"   - Graceful shutdown mechanisms are applicable to removal a TE
   resource (e.g., link, a group of TE Links or an entire node). Trigger
   for the graceful shutdown event is a local matter at the node
   initiating the graceful shutdown. Typically graceful shutdown is
   triggered for administrative reasons, such as
link maintenance or
   software/hardware upgrade at a node.
  "
 
 
In these cases, we have put a lot of effort into GMPLS to ensure that the LSP is capable of continuing to exist even when the control plane is down. 
 
Certainly, but as you know these are for unplanned outages, aka graceful restart (GR) mechanisms. This ID talks about planned outages: Graceful Shutdown (GS), like to one you mentioned above.
 
I am asking whether we want to give the ingress due warning that the LSP is about to enter this state (i.e. data plane up, but control plane disconnected). The reason why the ingress might be interested is that, once in this state, the LSP is not fully manageable.
 
An option is to provide an error value that says "Data plane OK, but control plane about to go away". This would allow the ingress to re-route (mb4b) if it cared. 
 
The ID assumes that Graceful Shutdown mechanisms are agnostic to control vs. data plan reasons for the planned outage. IMO we don't need specific codes for one vs. the other failure. Do you or anyone else feels that we need to have different cause code for the data vs. control plane failures (for e.g., some admin reasons).   
 
So the question for you and the WG is: Is this an important case?  
 
Yes, as mentioned above.
 
 Does the ingress ever care about this loss of control plane connectivity? 
Yes, for the reasons you listed above, e.g., software upgrades.
 
 
Cheers,
Adrian