[Date Prev][Date Next] [Thread Prev][Thread Next]
Re: EAP6 clustering
- From: Chhi'mèd Künzang <clkunzang gmail com>
- To: William DeCoste <wdecoste redhat com>
- Cc: users lists openshift redhat com, Jason Shepherd <jshepher redhat com>
- Subject: Re: EAP6 clustering
- Date: Wed, 19 Sep 2012 17:32:55 -0700
Thanks for the response, Bill. That explains everything and is good news.
I would vote for the automatic migration -- not only because it would be easier for me, but because even in the simplest case I think the process of switching applications is error-prone. Maybe in this case I can use a snapshot (need to investigate more) to make it simpler, but whenever manual intervention and judgment are required mistakes can happen.
For example, even though doing it was pretty straightforward, and I knew what needed to be done. I managed to resolve my merge conflicts wrongly (selected the new skeleton files rather than my old repository ones) when merging after creating my scalable app. Again, I know that was my mistake, and I was able to fix it -- but I think there is enough value in having the process require zero operator intervention in the case of existing apps.
So maybe consider that as a feature for future upgrades, even if it's not appropriate or practical do it in the present case. If you think about the situation of an established app relying on your platform, it's obviously going to be important to be able to apply bug fixes, and even upgrade to newer supported releases that are source compatible with existing apps. Having a guaranteed error-free way to do this would be valuable.
On Sep 19, 2012, at 5:24 PM, William DeCoste wrote:
> Hi Chhi'mèd,
> We've had a recent regression bug (https://bugzilla.redhat.com/show_bug.cgi?id=857205). For a little background ...
> 1) Previously when you created a scaled AS7/EAP6 application you got one gear for haproxy and one gear for JBoss. As the app scaled up and down, the JBoss instances would cluster
> 2) We recently added a JBoss instance to the haproxy gear so before any auto-scaling there are 2 JBoss instances running - one on the haproxy gear and one on a second gear). The haproxy JBoss will not cluster with the other JBoss instances - this is the crux of the bug above.
> The bug has been fixed and is in QA. The fix should roll into production in ~3 weeks. Existing applications will need to be migrated - we're undecided on wheterh we will do an automatic migration or if users will need to create a new JBoss app.
> I hope this helps. If you have any additional questions please let me know.
> Thanks -Bill
> On 9/19/12 5:08 PM, Chhi'mèd Künzang wrote:
>> Thanks, Jason. I saw your forum reply, and it looks like this is an acknowledged bug. As long as I can reasonably expect clustering to work eventually, then I am fine at my current stage of development. My initial correspondence with Nam left me unsure whether this was a simple bug or a deeper problem or policy decision.
>> Just to double-check: am I correct in assuming that you fully intend to support automatic EAP6 clustering for scalable apps, and that this is just a temporary hiccup?
>> On Sep 19, 2012, at 4:45 PM, Jason Shepherd wrote:
>>> Sorry for the delay in getting back to you. I was hoping we would have this issue sorted out, and then I would replay and ask you to try again. However the issue is still outstanding, and should be fixed soon.
>>> I am going to CC Bill DeCoste as well. Bill, can you provide any more details on whether this will work with TorqueBox, and what will be required to pick up the change? I noticed you said it would require a migration, does that mean users need to migrate to a new version of the JBoss cartridge?
>>> Jason Shepherd
>>> On 09/20/2012 07:42 AM, Chhi'mèd Künzang wrote:
>>>> Hi, I'm reposting a question I asked on the forum in case I might get a better response here. The original post was a week ago with no response yet.
>>>> Although I mention TorqueBox in the question, as far as I understand it, my question is applicable to all EAP6 applications.
>>>> I am trying to understand the issues around using OpenShift to cluster a TorqueBox application. TorqueBox, a Red Hat product, is a Ruby (via JRuby) application server built on top of JBoss.
>>>> I was told by a TorqueBox dev that OpenShift supports clustering for scalable applications. That would be great and one of the significant benefits of using OpenShift versus other options. However, when I emailed to inquire about this, I was told that EAP6 cannot be clustered even in a scalable application.
>>>> I was asked to start this thread and told that clarification would be provided here. I am quite interested in any information anyone can provide about what is currently possible and/or planned with regard to JBoss clustering. My goal is to run a TorqueBox application on OpenShift and take advantage of clustering when the application scales.
>>>> Thanks for any information you can provide.
>>>> users mailing list
>>>> users lists openshift redhat com
> Bill DeCoste
> Principal Software Engineer, Red Hat
> wdecoste redhat com
[Date Prev][Date Next] [Thread Prev][Thread Next]