[infinispan-issues] [JBoss JIRA] (ISPN-928) Interceptor that allows invocations only when cluster is formed of N nodes

[Tristan Tarrant (JIRA)]

     [ https://issues.jboss.org/browse/ISPN-928?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Tristan Tarrant closed ISPN-928.
--------------------------------
    Fix Version/s: 8.2.0.Final
       Resolution: Duplicate Issue


> Interceptor that allows invocations only when cluster is formed of N nodes
> --------------------------------------------------------------------------
>
>                 Key: ISPN-928
>                 URL: https://issues.jboss.org/browse/ISPN-928
>             Project: Infinispan
>          Issue Type: Feature Request
>          Components: Configuration, Core
>            Reporter: Galder Zamarreño
>            Assignee: Galder Zamarreño
>              Labels: hackathon
>             Fix For: 8.2.0.Final
>
>
> Following from https://github.com/pmuir/infinispan-examples/commit/f5d090092fa7b3660025b571af29882c720bb987#L6R12
> It'd be great to have an configurable StrictCluster interceptor in Infinispan which would basically make all invocations wait until the cluster of N nodes has been formed. I think it'd be a great addition and would allow clients to verify whether the cluster actually forms without the need to verify whether data replicates...etc.
> In principle, the configuration would be at the CacheManager, i.e.:
> <transport strictNumMembers="4"... />
> However, it could also be useful to configure it at the cache level. So, could maybe want to do this: I want cache X to allow invocations the moment I have 2 nodes (in spite of the cluster being formed of 4 noes), whereas I want cache Y to allow invocations once I have 3 nodes.
> Apart from an strict number of nodes, you could have a minimum number of nodes: allow invocations once I have 4 or more nodes. The strict value could still be useful to make sure intrusive machines don't get into the cluster, i.e. I expect 4 nodes in the cluster and if I have 5, something is wrong.
> I think it's an interesting concept that would get rid of cluster validation code in examples and RadarGun.



--
This message was sent by Atlassian JIRA
(v6.4.11#64026)