We are using eXtreme scale version 220.127.116.11 in Equinox environment.
1. What would happen if the primary and all (sync and async) replica shards lost?
2. How MMR would behave in the case of primary and replica shards lost?
Your insight would be highly appreciated.
Thank you in advance,
Notice: We have upgraded developerWorks Community to the latest version of IBM Connections. For more information, read our upgrade FAQ.
Pinned topic MMR sync when primary and replica shards have been losted
Answered question This question has been answered.
Unanswered question This question has not been answered yet.
Updated on 2012-11-13T16:26:58Z at 2012-11-13T16:26:58Z by andymc_
andymc_ 060000RAG316 Posts
Re: MMR sync when primary and replica shards have been losted2012-11-13T16:26:58ZThis is the accepted answer. This is the accepted answer.Hi Erwin,
1. This is a catastrophic failure. In this case, all data in the grid would be lost. MMR can help in this case, as can using a secondary data store (like a relational database - with a Loader).
2. MMR allows you configure multiple domains (essentially separate grids) that mirror each other. If a domain fails, it will be able to restore its data from another domain when it comes back up. However, the replication of updates between MMR domains is asynchronous, so there are times where the domains are in an inconsistent state (i.e. updates to domain 1 have not yet been replicated to domain 2). A catastrophic failure of an entire domain while it is still in an inconsistent state will result in some data loss (i.e. domain 1 fails before it is has replicated some updates to domain 2 - those updates are lost).
The real intent of MMR is to handle network "splits" - for example, suppose you have two data centers each with its own domain. Clients access grid data in both domains. Now, lets say that a network outage prevents the two data centers from communicating. Clients that can still access each domain will be able to continue working. Once the network between the data centers has been restored, the updates that occurred during the outage will replicate so that the data in each domains are eventually consistent.
Hope this helps,