Hide Forgot
For RHEL 6.3, we will need to document Redundant Ring Protocol configuration in the chapter on the cluster.conf file, if the feature makes the release. This is an email exchange I had with Fabio DiNitto on the background: On 2/8/2012 8:28 PM, Steven Levine wrote: > > Fabio: > > > > I'm looking at BZ#733298, which concerns support for Redundant Ring > > Protocol in RHEL 6.3, and trying to determine how to approach the > > documentation for this feature in the Cluster Administration Guide. I'm > > hoping you can help and advise me here. > > > > My first question: Will the annotated schema for the cluster.conf file > > be updated to include definitions of the new components and their > > options (altname, altmultcast)? If you are referring to the cluster relaxng schema, that´s already done. > > If so, I can use those definitions in > > the updated documentation for luci and ccs. luci and ccs have their BZ clone to deal with it. Ryan and Chris should be able to give you more precise update on their status. > > > > But other than just defining the parameters you can configure (with luci > > or ccs), do you think I should include some general overview of the > > feature, along the lines of the following: > > > > "For the RHEL 6.3 release, the Red Hat High Availability Add-On supports > > Redundant Ring Protocol, which allows you to configure your cluster with > > multiple redundant local-area networks..." > > > > And then maybe an overview of what you can do, and a list of what isn't > > supported (no mixing of IPv4 and IPv6, for example). Is there a writeup > > along these lines that could go in the document? Yes I think a general overview is a good idea. Just to make a few bits clear, before you dive into hours of work. 1) cluster/cman only configures corosync RRP based on corosync specifications. 2) at this specific point in time, RRP support is at risk due to a blocker bug in kernel. As for #1, the only documentation I have are the test results in that bugzilla that shows all valid and invalid combination for the configuration bits (see comment #3) and that gives you a basic idea of what´s supported or not. I think Steven or Jan can help you with a more descriptive text of the feature since they wrote it for corosync :) > > > > The major question, though, is whether I should add a section to the > > chapter on the cluster.conf configuration in the Cluster Administration > > document on how to configure RRP. BZ#733298 provides some examples, but > > I'm not sure which of those examples might go in the document. Is this > > something you can help me with -- if you think this should be documented? I think that documentation is good, but I would wait for #2 to be resolved before spending hours and hours on it. The examples with RRP can become rather difficult to explain in "end user terms". > > > > I know this is a pretty large request, perhaps, but at this point I'm > > looking to see what we should provide in general. > > > > Thanks for any help or pointers to existing documentation, Yup, I´ll be very happy to help, once we know this is going to make 6.3. Otherwise i´d prefer to postpone the work because I am rather overloaded with 7 at the moment. Fabio
Moving to 6.4: Let's not document RRP now. We'll will work with our one customer to teach them how to use it. Sayan On 03/13/2012 12:11 PM, Steven Levine wrote: > At this point, based on my understanding of where RRP support stands, I > would like to confirm and clarify the documentation plans for how to > address (or not address) this feature in RHEL 6.3. > > Although there are 2 components to this feature (corosync and DLM over > SCTP), my understanding after talking with Dave Teigland is that you > would really need to use both components for the feature to be useful. > This would mean that even if the corosync aspect is supported, if the > DLM over SCTP aspect is in technical preview then RRP support as a whole > would remain in technical preview. Is that correct? > > Or, specifically, do I document RRP configuration (with luci, ccs, and > the cluster.conf file) in RHEL 6.3? > > In the past, in cluster documentation, we have not documented features > in tech. preview in the regular documentation. Those features are > documented in the release notes and called out as being in technical > preview, but once they are in the regular documentation set on > redhat.com they imply some level of promise that there will be full > support in the next (or a near future) release. Those released documents > have a life of their own. It seems a bad idea to document a > not-yet-supported feature that we may withdraw, in the standard document > set. > > So even though Conga in 6.3 shows a configuration screen for RRP > components, and the ccs command supports this as well, I am currently > thinking that I should not call out this feature in the 6.3 Cluster > Administration document (or suggest that John Ha add an overview of the > feature to the HA Add-On Overview). > > Does this make sense? > > -Steven > > > > > > > > > > > >