[Pgpool-general] suggestion to difference detection engine

[Tatsuo Ishii]

> Greetings,
> 
> writer ----> table <-----> reader (gets only the first new record for
> processing)
> 
> In my test setup one thread polls a table for new records while
> another thread writes new records in it.  I get "pool_process_query:
> kind does not match between backends master ..." errors after a few
> minutes.  Probably the reader thread has seen a new record that is
> already in the primary but is not yet in the secondary.  Is there a
> way to fix this problem.  Perhaps if pgpool will be able to make a
> number of retry attemps and if afterwards the results are still
> different then only will pgpool stop replication.  I wonder if this is
> possible.

That'll be pretty tough since the retry might not never end. Do you
have only one writer/reader? Then you could have an exclusive table
lock on the table so that the reader never reads half replicated data
from the table.
--
Tatsuo Ishii
SRA OSS, Inc.