You are right, the transactions are only persistent on a single server -
if that server dies, you will loose the transaction. So you cannot
ensure transaction failover.
However, you can still use the clusterer module to do dialog failover,
since dialog is replicated across all nodes in the cluster. Hence if one
goes down, you will have the data on all other nodes, and if a
sequential message reaches any of those servers, they will be able to
route it properly to the endpoint.
Hope this helps.
On 10/16/18 5:32 PM, Igor Olhovskiy wrote:
> Wanted to clarify one topic from clusterer module
> According to article
> https://blog.opensips.org/2018/03/21/full-anycast-support-in-opensips-2-4/ > I’ve got, that transactions are really persist only on one server.
> So, if packet for some reason comes to other server, lookup on other
> nodes would be preformed.
> But if node, that holds info especially for current transaction(or
> dialog) will die in a middle of process?
> Question is, can I use clusterer module for failover as well?
> (Best scenario also would not involve any SQL database for persistent
> storage, maybe only redis)
> Regards, Igor
> Users mailing list
> [hidden email] > http://lists.opensips.org/cgi-bin/mailman/listinfo/users >