I think it makes sense not to validate store configuration unless we know
that the entry is enlisted as WRITE.

I've created the issue: https://issues.apache.org/jira/browse/IGNITE-3086

2016-05-04 5:28 GMT-07:00 Denis Magda <[email protected]>:

> As Val already mentioned you can't mix write-through and write-behind
> caches
> in a single transactions.
>
> However, in my understanding I don't see any reason why we can't mix these
> caches taking into account that the one will be used for 'read' operations
> only.
>
> Alex G. or Sam can you share your thoughts? Do we need to create a JIRA
> ticket?
>
> In any case presently I see the following solutions:
>
> 1) make all caches that participate in the transaction either write-through
> or write-behind;
>
> 2) use IgniteCache.getAllOutTxAsync(...) for the cache that is used to read
> data, probably it will work in your case. Note that this should be used
> only
> if you don't want to acquire locks during the read operation from the
> cache.
>
> 3) use READ_COMMITTED isolation level instead of REPEATABLE_READ (default
> one). However, the lock won't be acquired on the 'get' operation as in case
> 2.
>
> --
> Denis
>
>
>
> --
> View this message in context:
> http://apache-ignite-users.70518.x6.nabble.com/Error-on-Transactions-with-Write-Behind-tp4736p4764.html
> Sent from the Apache Ignite Users mailing list archive at Nabble.com.
>

Reply via email to