You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@ignite.apache.org by xmw45688 <xw...@procurant.com> on 2019/02/27 07:31:36 UTC

Re: same cache cannot update twice in one transaction

It seems that this enhancement has not been implemented yet for the following
cases:
trx.start() {
1. update t1 set col1='a' where col2='c';
2. update the same table t1 with cache API. 

}
trx.end();

Can someone confirm?  
many thanks, Xinmin



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

Re: same cache cannot update twice in one transaction

Posted by xmw45688 <xw...@procurant.com>.
1. why do we need the storage (whether native or 3rd party)?  I need the
storage because 1) down time for production deployment. 2) the data in the
cache  may be crashed and All ignite clusters are completed down (sorry)

2. So I need a storage to store the  data that have been committed when
transactions are committed.  Do I have to care MVCC query in 3rd party
storage? No.  It's very difficult (if not possible) to achieve this goal
given people are using distributed NoSQL databases.   Can we directly query
Ignite native storage with MVCC without the caching layer running on top of
it?  

3. Comparing to the other database storage, Ignite native storage is
relative new.  The features to support ML, data mining are provided, but not
comprehensive like Apach Spark...  So questions are raised when the
integration of Ignite native storage or 3rd party storage with Spark - which
storage is better?

4. I think that Ignite should specialize in caching, MVCC on cache layer
(grid). Provide own native storage, ML, data mining feature are great, but
there is limited resource for this community to complete this mission. 
Leave the other specialized groups (exports) focusing on storage store, ML,
data mining, Block Chain...  

5.  Instead Ignite might focus on to integrate these things together to
build data-ecosystem.  I think that there are many dis-integrated, separated
data storage, cache, ML, Deep Learning, blockchain. To put these things
together, it needs visionary leaders, high skilled engineers.  Why can we
make our world life easier? 

6. Igniters, please focus the grid computing (cache), ACID-compliant
transaction, MVCC-enabled SQL in cache layer, and provide services to
integrate with the other storage, spark (ML), hyper ledge fabric or ethereum
or bitcoin (block chain).  



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

Re: same cache cannot update twice in one transaction

Posted by Павлухин Иван <vo...@gmail.com>.
Hi,

MVCC in Ignite is targeted to provide transactional consistency
guarantees. I suppose that with eventually consistent 3rd party store
it would be impossible to give much guarantees in general. Do think
that such eventually consistent store will be widely used? What kind
of guarantees should it provide? Is it easy to use it properly?
Currently we do not have answers to these questions. Feedback is
appreciated.

Also I must say that MVCC feature is currently in beta stage and
limitations are listed in documentation [1].

[1] https://apacheignite.readme.io/docs/multiversion-concurrency-control#section-other-limitations

чт, 28 февр. 2019 г. в 22:34, xmw45688 <xw...@procurant.com>:
>
> Hi Ilya,
>
> It'd better if this was mentioned in Ignite Doc.
>
> It seems very limited if MVCC only supports  the Ignite native persistence.
> Yes, supporting MVCC in 3rd party persistence is challenging.  However, do
> we really need MVCC when the data from Cache (where MVCC already enabled) is
> ready to write to a 3rd party persistence store.   I think that an "eventual
> consistence" for writing cached data into a 3rd persistence layer seems
> sufficient when Ignite is used as cache stored, and the data in cache store
> is persistent.
>
> Does Ignite have a plan to support MVCC in cache layer and write the data
> from the cached store into a 3rd party persistence store with some limited
> feature like "eventual consistence".
>
> Can some gurus shed some lights on this subject?
>
>
>
>
>
> --
> Sent from: http://apache-ignite-users.70518.x6.nabble.com/



-- 
Best regards,
Ivan Pavlukhin

Re: same cache cannot update twice in one transaction

Posted by xmw45688 <xw...@procurant.com>.
Hi Ilya, 

It'd better if this was mentioned in Ignite Doc.  

It seems very limited if MVCC only supports  the Ignite native persistence. 
Yes, supporting MVCC in 3rd party persistence is challenging.  However, do
we really need MVCC when the data from Cache (where MVCC already enabled) is
ready to write to a 3rd party persistence store.   I think that an "eventual
consistence" for writing cached data into a 3rd persistence layer seems
sufficient when Ignite is used as cache stored, and the data in cache store
is persistent.  

Does Ignite have a plan to support MVCC in cache layer and write the data
from the cached store into a 3rd party persistence store with some limited
feature like "eventual consistence".  

Can some gurus shed some lights on this subject?





--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

Re: same cache cannot update twice in one transaction

Posted by Ilya Kasnacheev <il...@gmail.com>.
Hello!

Yes, unfortunately MVCC does not support 3rd party persistence (and I'm not
sure that it is possible by design).

Regards,
-- 
Ilya Kasnacheev


чт, 28 февр. 2019 г. в 00:29, xmw45688 <xw...@procurant.com>:

> Hi Ilya,
>
> Since I'm using Cassandra as data store, it raises the following exceptions
> once MVCC is enabled -
> <property name="atomicityMode" value="TRANSACTIONAL_SNAPSHOT"/>
>
>
> class org.apache.ignite.IgniteCheckedException: Grid configuration
> parameter
> invalid: readThrough cannot be used with TRANSACTIONAL_SNAPSHOT atomicity
> mode
>         at
>
> org.apache.ignite.internal.processors.GridProcessorAdapter.assertParameter(GridProcessorAdapter.java:140)
>         at
>
> org.apache.ignite.internal.processors.cache.GridCacheProcessor.validate(GridCacheProcessor.java:527)
>         at
>
> org.apache.ignite.internal.processors.cache.GridCacheProcessor.createCacheContext(GridCacheProcessor.java:1543)
>         at
>
> org.apache.ignite.internal.processors.cache.GridCacheProcessor.prepareCacheContext(GridCacheProcessor.java:2324)
>         at
>
> org.apache.ignite.internal.processors.cache.GridCacheProcessor.lambda$null$fd62dedb$1(GridCacheProcessor.java:2163)
>         at
>
> org.apache.ignite.internal.processors.cache.GridCacheProcessor.lambda$prepareStartCaches$5(GridCacheProcessor.java:2086)
>         at
>
> org.apache.ignite.internal.processors.cache.GridCacheProcessor.lambda$prepareStartCaches$937cbe24$1(GridCacheProcessor.java:2160)
>
>
>
>
> --
> Sent from: http://apache-ignite-users.70518.x6.nabble.com/
>

Re: same cache cannot update twice in one transaction

Posted by xmw45688 <xw...@procurant.com>.
Hi Ilya,

Since I'm using Cassandra as data store, it raises the following exceptions
once MVCC is enabled - 
<property name="atomicityMode" value="TRANSACTIONAL_SNAPSHOT"/>


class org.apache.ignite.IgniteCheckedException: Grid configuration parameter
invalid: readThrough cannot be used with TRANSACTIONAL_SNAPSHOT atomicity
mode
	at
org.apache.ignite.internal.processors.GridProcessorAdapter.assertParameter(GridProcessorAdapter.java:140)
	at
org.apache.ignite.internal.processors.cache.GridCacheProcessor.validate(GridCacheProcessor.java:527)
	at
org.apache.ignite.internal.processors.cache.GridCacheProcessor.createCacheContext(GridCacheProcessor.java:1543)
	at
org.apache.ignite.internal.processors.cache.GridCacheProcessor.prepareCacheContext(GridCacheProcessor.java:2324)
	at
org.apache.ignite.internal.processors.cache.GridCacheProcessor.lambda$null$fd62dedb$1(GridCacheProcessor.java:2163)
	at
org.apache.ignite.internal.processors.cache.GridCacheProcessor.lambda$prepareStartCaches$5(GridCacheProcessor.java:2086)
	at
org.apache.ignite.internal.processors.cache.GridCacheProcessor.lambda$prepareStartCaches$937cbe24$1(GridCacheProcessor.java:2160)




--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

Re: same cache cannot update twice in one transaction

Posted by Ilya Kasnacheev <il...@gmail.com>.
Hello!

Are you sure that you have enabled MVCC for your cache?

Regards,
-- 
Ilya Kasnacheev


ср, 27 февр. 2019 г. в 10:31, xmw45688 <xw...@procurant.com>:

> It seems that this enhancement has not been implemented yet for the
> following
> cases:
> trx.start() {
> 1. update t1 set col1='a' where col2='c';
> 2. update the same table t1 with cache API.
>
> }
> trx.end();
>
> Can someone confirm?
> many thanks, Xinmin
>
>
>
> --
> Sent from: http://apache-ignite-users.70518.x6.nabble.com/
>