Re: [infinispan-dev] upgrade to 5.2.x
by Ales Justin
>> I haven't actually used infinispan-query, so I don't know if
>> infinispan-core needs to have its classes visible. Can you comment?
>>
>>
>> https://github.com/capedwarf/capedwarf-jboss-as/blob/master/build/src/mai...
>>
>>
>> Afaik, it's needed for Infinispan to pick-up Infinispan-Query's
>> extension/integration hooks;
>> e.g. org.infinispan.query.impl.LifecycleManager in META-INF/services
>> ServiceLoader pattern
>
> I think your going about it backwards, though. The infinispan-query module should depend on infinispan-core, not the other way around. When creating the GlobalConfiguration, specify the classloader of the infinispan-query module. You can do this through the subsystem configuration by specifying a module attribute within <cache-container/>.
Hmmm, OK, this might work if the Infinispan-Query is the only hook you want to get picked-up by the Infinispan-Core ServiceLoader pattern lookup.
It won't work if there are more.
What are other known, similar to Infinispan-Query - via ServiceLoader pattern, hooks in Infinispan?
Afaik, users can add their own?
-Ales
12 years, 7 months
Infinispan and CAP Theorem
by Israel Lacerra
Guys,
Sorry if this is not a "pure" dev question... but I guess the devs now
better how to answer this. If nobody wants to answer, no problem! :)
I writing about CAP and I will talk about how Infinispan deal with it. But,
I'm getting a little confused about this:
https://docs.jboss.org/author/display/ISPN/Consistency,+Availability+and+...
Reading the new Brewer article on infoQ (
http://www.infoq.com/articles/cap-twelve-years-later-how-the-rules-have-c...),
is my understand that the application can choose between A and C. Partition
will occur always, and it's not up to the application to decide this. Even
in a network without communication failures, when we writing in a node, if
someone write in another node "almost" at the same time, the system will let
the two writes happen besides the latency (partition) between the two nodes.
So, what is making me confuse is:
"Infinispan has traditionally been biased towards Consistency and
Availability, sacrificing Partition-tolerance"
I know about the async options of Infinispan. So, for me, the users can
decide at a certain degree of consistency x availability. But, how
Infinispan can sacrifice partition-tolerance?
cheers!
Israel
12 years, 7 months
Update on next release date (5.2.0.Alpha3)
by Mircea Markus
Hi,
The main goal of the next Alpha release is getting NBST(ISPN-1424) out (alpha-stable). After discussing with Dan and Adrian, here's the action plan:
- Dan/Adrian will issue a pull request Mon/Tue (6-7 Aug)
- given the complexity of the change, the review for this is expected to take up to a week. During this time, besides answering review comments, they will consolidate the implementation
- finally the release is estimated to happen at the beginning of the week starting 13 August - the day to be confirmed later on next week
Thanks,
Mircea
12 years, 7 months
Exposing numOwners through JMX
by David van Balen
After looking at the JMX component documentation (http://docs.jboss.org/infinispan/5.1/apidocs/jmxComponents.html) and browsing around in jconsole, I don't see the numOwners attribute for distributed caches exposed anywhere, although I thought I had seen it before.
In any case, the only way I see numOwners being retrieved in the ISPN source is by calling DistributionManager.locate for a particular key. Is there a way to retrieve the currently configured value of numOwners, regardless of whether it's the default or was overridden in the configuration?
12 years, 7 months
