Github user kapustor commented on the issue:
https://github.com/apache/incubator-hawq/pull/1353
Hi @sansanichfb!
> Also, I am wondering, which target database patch was tested against? The
transactional approach doesn't seem clear to me - if DB doesn't support
transactions - you are still sending data in batches, which might leave target
db in the non-consistent state since you can't rollback the last transaction.
Should we reconsider this approach and send only one batch(not ideal though,
maybe ask users to tweak partitioning) in that case?
We have tested the patch with Oracle, Postgres, MSSQL and Ignite cluster
targets - they all work fine, the perfomance is good enough (especially for
Ignite - about 200 Mbyte/s for two-node Greenplum and Ignite clusters).
If you know any other target that should be tested - pls let me know, I
will do it.
As for transactional approach - even if the target do support transactions,
there still may be inconsistense in data, for example if one of the GP
segements will fail to insert it's batch (other segements may succeed and only
part of the data will be inserted). As for sending all the data in one batch -
as I understand, such approach will limit the max data size by memory available
for PXF service, because PXF have to aggregate the batch in its memory.
---
