Is there a way to configure wildfire such that multiple instance become part of a single cluster as part of a load-balancing or fail-over design?
Raymond
Raymond,
This is not currently possible. However, it’‘s a feature we’'re targetting over the next several months. Most likely, it will be a Wildfire Enterprise feature, for a few reasons:
-
The clustering will likely use Tangosol’'s Coherence technology, which is commercial.
-
Fail-over is something we consider to be an “enterprise” feature. Scalability will always be a part of Open Source (which is why the connection manager feature in 3.0 was built as open source), but being able to fail-over between servers is a mission critical feature commercial feature.
For the time being, we’'d recommend using a warm back-up server. You could also use a replicated database setup. With that combination, you should never have data integrity issues, and downtime would be quite low.
Hope that helps!
Regards,
Matt
perhaps its something that could be built (partially) into Spark? i.e. if it can’'t connect to one server it goes for the other.
I know this isn’'t real failover etc, but would at least keep people talking…
I.
btw - I know thats not a Wildfire issue, but since its in here thought I’'d mention it.
Matt,
- The clustering will likely use Tangosol’'s Coherence technology, which is commercial.
Out of curiosity, what are the crutial features in Tangosol’‘s Coherence clustering suite that are so compelling that JiveSoftware can’'t apt for other clustering suites (like JINI maybe)?
- Fail-over is something we consider to be an “enterprise” feature. Scalability will always be a part of Open Source (which is why the connection manager feature in 3.0 was built as open source), but being able to fail-over between servers is a mission critical feature commercial feature.
That would certainly cause Wildfire to loose some of its votes over eJabberd which is enterprise class despite being free. I could understand the commercial factors behind the decision. But IMHO, that should not be the modus operandi. I’‘d prefer to see Wildfire become the “de facto” XMPP server and make virtually everybody churn out from using other flavors of XMPP servers. You may loose some by making the clustering part Open Source. However, my naive observation tells me that because it’‘s CLUSTERING, it’‘s simply DIFFICULT even for the majority tech people. Many (companies especially) can’‘t afford to rely on open community for support. That’‘s when the commercial support sells. A good example might be the RedHat Clustering Suite & GFS. I believe you’'re in the better position to come out with a handful list of examples. In other words, the more open it could be, the more people will try (no expiry date, please), the more they get confortable, the more they gain confidence, the more they use, and the more complicated their system becomes, the harder to resist buying for commecial support.
Anyway, you do have strong reasons to justify your point don’‘t you? It’'s just my 10 cent opinion
p/s: Can’'t resist the ease provided by Wildfire administration console and extensibility via plugin
aznidin wrote:
I’'d prefer to see Wildfire become the “de facto” XMPP server and make virtually everybody churn out from using other flavors of XMPP servers.
Well, I am from the ejabberd team and I prefer to see XMPP servers to become the “de facto” IM servers and make virtually everybody churn out from using other flavors of IM servers. Regarding Wildfire become the “de facto”: I do not believe that will happen, I even think that after some time jabberd14 might take back some market share of jabberd2, ejabberd and Wildfire Server But the most important thing is that market share from proprietary networks drops in favour of Jabber servers
CLUSTERING, it’'s simply DIFFICULT even for the majority tech people.
Well, that is a relative thing: there are even people that think that Linux is more difficult than Windows, while I just think the reverse; I even think FreeBSD is more userfriendly than Windows :o)
Hi Sanderd,
Thanks for sharing your view.
sanderd wrote:
Well, I am from the ejabberd team and I prefer to see XMPP servers to become the “de facto” IM servers and make virtually everybody churn out from using other flavors of IM servers.
Great! I’‘d like to share with you my story. I was at a time, very convinced to choose ejabberd as my platform. The clustering + all the enterprise features were just what I want. But when I thought back about it, in the long run, it’‘s gonna kill me. Given the team that I have here, ejabberd’‘s admin console is scary. Everything (problem, support, maintenance, etc…) will then fall back to me. With Wildfire admin console, I have a broader choice of who could use it. It will be perfect if I could have ejabberd’‘s enterprise features and Wildfire’'s admin console.
Guys, pls accept my appologies for telling my ejabberd story here
sanderd wrote:
Regarding Wildfire become the “de facto”: I do not believe that will happen
Guess what? Me neither , though it’‘s not impossible. What’‘s important is to maintain the healthy competition. I was talking about if I were in Jive’‘s shoes. Then I’'d like to see Wildfire wins. And clustering solution is the key.
sanderd wrote:
But the most important thing is that market share from proprietary networks drops in favour of Jabber servers
Can’'t agree more with you on this :).
sanderd wrote:
Well, that is a relative thing: there are even people that think that Linux is more difficult than Windows, while I just think the reverse; I even think FreeBSD is more userfriendly than Windows :o)
True, it’‘s relative. But the fact that clustering is difficult might simply imply why it’'s so expensive even for its support