Tips for high availability ticket registry?

[John Rellis]

Hello,

I am deploying the cas server on two servers running tomcat that are behind an apache load balancer.

Does anyone have any opinions on which cache provider is easiest and or best to use for the HA ticket registry???  

The docs mention the use of Hazelcast, EhCache, MemCached, Ignite or Couchbase.  I haven't really used any of these to such an extent that I'd be leaning to favour any so pros or cons would be greatly appreciated.

Thanks,

John

[John Rellis]

I was also curious, i was able to get the JPA ticket registry working (that is, storing the tickets in the database) in a few minutes.  Is there any reason not to use this in a load balanced scenario if there's a highly available database server that is easily reachable?

We are looking at users in the 100s not in the 1000s.

[Dmitriy Kopylenko]

This is an old thread, but might answer some questions for you of why RDBMS-backed ticket registry impl might not be the best choice: http://jasig.275507.n4.nabble.com/JpaTicketRegistry-A-Sinking-Ship-td4256973.html

Best,

D.

-- 
You received this message because you are subscribed to the Google Groups "CAS Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cas-user+u...@apereo.org.
To post to this group, send email to cas-...@apereo.org.
Visit this group at https://groups.google.com/a/apereo.org/group/cas-user/.
To view this discussion on the web visit https://groups.google.com/a/apereo.org/d/msgid/cas-user/7bae6bd6-d169-4aff-8b56-6ef4c05f6492%40apereo.org.
For more options, visit https://groups.google.com/a/apereo.org/d/optout.

[John Rellis]

Excellent, thank you!

[Dmitriy Kopylenko]

Also to answer your previous question of “what is the best distributed ticket registry implementation” - no one could give this type of an answer (“the best”, that is). These days, the Hazelcast ticket registry seems to be the very popular one among other “in-memory” impls, and seems to be the easiest to set up and (from personal experience). In CAS v 4.2.x+ for example, in order to set up highly available cluster of CAS nodes with Hazelcast ticket registry, one only needs to include the module dependency in CAS’s overlay’s pom, add member nodes in the properties file and it just works!

But again, you have to see for yourself and try out all the available options to pick the one that you feel most comfortable with.

Best,

D.

To view this discussion on the web visit https://groups.google.com/a/apereo.org/d/msgid/cas-user/6B55CE13-3716-4532-9029-6DEE47CFFE2B%40unicon.net.

[John Rellis]

Thanks, yes you are right, it was too broad a question really, I suppose I was hoping for some "DEFINITELY DONT USE THIS ONE" answers to save some pain.  The link you provided is an old one alright and I wonder if anyone out there is using JPA on a 4+ release of cas (i'm using 4.2.2) with 100s of users that hasn't been awoken @ 3am due to the issues with it.  

From a reply offline and from googling around it seems people are having relatively low levels of pain with hazelcast.  I'll give it some time tomorrow to see if I can get it working and make a decision from there I think.

I was leaning towards JPA as the dev and support team will be able to debug the database easier, purely due to familiarisation.  since we are not yet living the dream of containerisation, distributed caches could cause operational headaches too as it is yet another component to manage in production.  

Thanks :)

[John Rellis]

Thanks folks, you are all very nice people, hazelcast "just worked" from what I can gather, nice work cas team.

Thanks all for your input!