> Hello,
> Yes, tried that, but did not solve it. Only the cited approach did it.
> We even tried the same idea: explicitly locking the entire table during 
> inserts,
> updates, deletes. The results were as expected at performance (VERY poor as
> pgpool "serialized" all these previously simultaneus transactions), and an
> unexpected side effect (with our confs): RAM usage "exploded" during stress 
> test
> and we had to hard reboot the machine.
> The kind of transaction row locking was the less penalty we found.
> 
> We are still stress loading the pool and it did not have a single failure 
> since
> then.
> It worked for our php app, but it should be tested for other apps. 
> How about binary only unmodifiable apps?

Ok, so your application issues not only INSERT but other DMLS. Or
worse, DMSs use values from prevous SELECT. Anyway, problem is I don't
know what are your SQLs which cause your problem. If you could show us
a self contained test case, I may be able to think of better way...
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp
_______________________________________________
Pgpool-general mailing list
[email protected]
http://pgfoundry.org/mailman/listinfo/pgpool-general

Reply via email to