LAE

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: alyon

Hi,

I had posted following query on another sub forum, but did not get any response.

Hoping to get the response here.
__________________________________________________ ________
One client have asked me if we can run multiple instances of same graph concurrently.

That graph queries and modifies few DB tables
I am not sure if brain DB nodes have the capability of waiting for the tables to be released? Can they run into some kind of race condition and hence produce wrong results?

Any ideas?
__________________________________________________ __________
Thanks
Anshul

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: rpigneri

Dear Anshul,

Sorry to hear that you did not get an answer in the other forum. Could you tell us where that is to see where we should put the final answer?

I do not have much visibility into the DB nodes, but if there is a way to guarantee that these nodes fail when the database is locked, you could use the ls.brain.controller.restartMaximum and ls.brain.controller.restartableErrors parameters to restart the entire graph when one of these race conditions occurs. However, that would require you to restart the entire graph on all restartableErrors and that the graph is started with some parameters on the controller. Not certain if that is a workable solution for you, but this might get you halfway there.

Hope that helps,

Rocco

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: alyon

Hi Rocco,

Thanks for your reply.

The graph in question is not suitable for restart, as it is an archive process.
Restarting will leave finles hanging or corrupt.

Anyway,
It was not the original requirement to be able to run in parallel.
Client simply wanted to know if the parallel run of the graph is possible or not.
It will save them some time while trying to process multiple files.


It will anyway be intersting to know if bre has the support for running same DB nodes in parallel or not.

Regards
Anshul

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: rpigneri

Dear Anshul,

I have done a bit more research into this situation and found that the LAE Database nodes do support the parallelism you are seeking.

When querying tables and executing SQL, the database simply handles the concurrency for you. However, the loading side is a bit more complex. Specifically, if you are entering data into an Oracle database, you can set the ParallelLoad option to true in order to enable parallel loading. However, there are some limitations on what type of table you can load data into using the ParallelLoad feature. This option uses direct path loading so it is subject to the same limitations as using direct path loading in SQL*Loader. Just be sure that you have LoadMethod set to Optimized in order to use the ParallelLoad option.

Both of these options are available on the Parameters II tab of the DB Store node.

Hope that helps,

Rocco

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: alyon

Thanks Rocco...