On Sat, May 25, 2013 at 11:49 AM, Afkham Azeez <[email protected]> wrote:
> > > On Sat, May 25, 2013 at 11:37 AM, Supun Malinga <[email protected]> wrote: > >> -- >> Supun Malinga >> Sent from my phone. >> >> On 24 May 2013 02:37, "Afkham Azeez" <[email protected]> wrote: >> > >> > There is no guarantee that the cluster message will be reliably >> delivered to all members. For instance, due to a momentary network delay or >> delay in the node, other members may detect that a member has left, but in >> reality, that member may be still running. If a cluster message, such as a >> DepSync message is sent in such a scenario, the node will not receive that >> message, and may never update its deployment repo. >> >> Hi azeez, >> >> I understand the requirement and the need. But if the clustering has >> failed(member dropped) means that node will no longer receive load from the >> load balancer right? >> > > There can be momentary failures due to delays in the nodes or network. > When this happens, nodes will detect that the member has left. However, > later, when the nodes become reachable again, they will be added back to > the cluster. They would have missed message which were received during the > time they were removed from the cluster. The solution would be to replay > selected messages from the past 5 minutes to any new node that joins. > +1. We can form a queue of such cluster messages once a node leaves the group and if the node joins the cluster again within a particular time period, we can do a 'replay' from the queue of messages that we have for the node. If the node does not join during this time period, we can discard the message queue. This is better than polling IMHO, say periodic polling for updates in worker nodes. > > >> > >> > We could implement one of the following solutions; >> > 1. For selected cluster messages, we could replay them if a new member >> becomes available. We can't blindly replay all messages, but replaying the >> depsync message will cause no harm because the depsync operation is >> idempotent >> > >> > 2. For depsync, we can have periodic polling, but the polling interval >> should vary from about 1 minute to 5 minutes. >> > >> > -- >> > Afkham Azeez >> > Director of Architecture; WSO2, Inc.; http://wso2.com >> > Member; Apache Software Foundation; http://www.apache.org/ >> > >> > email: [email protected] cell: +94 77 3320919 >> > blog: http://blog.afkham.org >> > twitter: http://twitter.com/afkham_azeez >> > linked-in: http://lk.linkedin.com/in/afkhamazeez >> > >> > Lean . Enterprise . Middleware >> > >> > _______________________________________________ >> > Dev mailing list >> > [email protected] >> > http://wso2.org/cgi-bin/mailman/listinfo/dev >> > >> >> > > > -- > *Afkham Azeez* > Director of Architecture; WSO2, Inc.; http://wso2.com > Member; Apache Software Foundation; http://www.apache.org/ > * <http://www.apache.org/>** > email: **[email protected]* <[email protected]>* cell: +94 77 3320919 > blog: **http://blog.afkham.org* <http://blog.afkham.org>* > twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez> > * > linked-in: **http://lk.linkedin.com/in/afkhamazeez* > * > * > *Lean . Enterprise . Middleware* > > _______________________________________________ > Dev mailing list > [email protected] > http://wso2.org/cgi-bin/mailman/listinfo/dev > > -- Thanks and Regards, Isuru H.
_______________________________________________ Dev mailing list [email protected] http://wso2.org/cgi-bin/mailman/listinfo/dev
