memcached failover

[Mark Cairney]

Hi,

I've no experience of using Hazelcast with Shibboleth but I've been
using it for our Openfire Jabber service for a while and it seems to be
pretty robust and handles nodes joining/leaving the pool sensibly.

My very limited understanding of memcached clustering was that it should
be doing lightweight replication i.e. every node connects to it's own
and everyone else's memcached so sessions should be written to all
memcaches.

Therefore I was a little bit puzzled as to why users were asked to log
in and it proved practically useless when it came to replicating consent
information (which we've disabled for now anyway).

Hopefully Hazelcast will provide the advantages Terracotta gave in this
respect but without the disadvantages :D


On 11/08/2016 03:33, Cantor, Scott wrote:
> On 8/10/16 8:52 PM, Paul B. Henson wrote:
>
>> But I've been using it in production for over two months and it's been
>> working great. JJ -- when are you going to polish this up and put together
>> a release :)? As cool as it is I feel a bit awkward with the "assemble
>> from parts" instructions ;).
> I don't know anything about Hazelcast, but if it works that well,
> probably the best option would be if JJ were willing to put it into a
> donateable form and be willing to act as a backstop supporting it if we
> included it in 3.3.
>
> I'm willing to spend a little time reviewing what it's doing to see if
> it's taking any shortcuts I would be uncomfortable with.
>
> -- Scott
>


-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.