1. As this doc says: https://apacheignite.readme.io/docs/3rd-party-store: Ignite writes through results of its SQL INSERT, UPDATE and MERGE queries. 2. Transaction in Ignite means absolutely the same as in another systems. If you want two or more changes to be in one transaction(i.e., classic example with transfer between bank accounts), then, for sure, you need to use transactions.
Evgenii 2018-07-09 16:52 GMT+03:00 Prasad Bhalerao <prasadbhalerao1...@gmail.com>: > Resending my last mail ... > > I understand that FULL_SYNC is required to keep the consistency between > primary data node and backup data node and not between the cache and 3rd > party data. > Sorry for asking the incorrect question. > > 1) Does ignite supports write through updates to 3rd party store when > using update or delete sql? > > 2) Is it mandatory to use ignite transaction to keep strong consistency > between in memory cache and 3rd party data store? > -Like you said, if I am updating more than one entry in cache then > transaction is necessary to keep strong consistency between 3rd party db > and ignite cache. Am I correct? > > Thanks, > Prasad > > On Mon, Jul 9, 2018 at 6:38 PM Evgenii Zhuravlev <e.zhuravlev...@gmail.com> > wrote: > >> Hi, >> >> 1. Cache write synchronization mode is not about 3rd party store, it's >> about waiting for write replies from other nodes. But Ignite with >> enabled 3rd party store guarantees consistency itself. >> >> 2. Ignite creates implicit transactions for atomic updates to guarantee >> consistency between nodes. If you want to make more than one update in >> transaction, then, of course, you should use explicit Ignite transactions. >> >> Evgenii >> >> 2018-07-09 15:37 GMT+03:00 Prasad Bhalerao <prasadbhalerao1...@gmail.com> >> : >> >>> Hi, >>> >>> I am using FULL_SYNC cache write synchronization mode. >>> 1) Does FULL_SYNC mode guarantees 100% consistency between ignite in >>> memory cache and 3rd party data store (oracle in my case)? >>> >>> 2) Does FULL_SYNC mode requires to do the cache updates in ignite >>> transaction?( I am using write through approach.) >>> >>> >>> Thanks, >>> Prasad >>> >> >>
