From: thomas at tada.se (Thomas Hallgren)
Date: Sat, 29 Oct 2005 10:35:54 +0200
Subject: R: Re: [Pljava-dev] crash due to huge load?
In-Reply-To: <43621A50000027BB@mail-5.mail.tiscali.sys>
References: <43621A50000027BB@mail-5.mail.tiscali.sys>
Message-ID: <4363346A.4000503@tada.se>


rcolmegna at tiscali.it wrote:
> Thomas,
>
> thanks for your reply. Sorry but I would disagree with your hipotesys 
> about Log4J: I have used it without any problem (in PLJava Trigger). 
>
> The evidence is the fact that if I use my PLJ/Trigger a lot of time (the
> same .SQL
> script execute, via PSQL, 20 times), without complex-thread behaviour, the
>
> Log4J logs are correctly produced.
>
>   
Yes of course log4j will produce output. That's not what I meant. What I 
mean is that:
- the java.util.logging package is well integrated with the PostgreSQL 
backend and you can control it using the log_min_messages, 
client_min_messages etc. in the postgresql.conf.
- the output will not conflict with the output from the PostgreSQL 
logger since one uses the other.
- by setting client_min_messages, you can actually propagate logging to 
the client.
- the java.util.logging does pretty much the same tasks as the log4j 
logger so you have a lot of redundant code in your backend for no 
apparent reason.
- if you want log4j to output to a file, you must use the untrusted 
language handler, the trusted handler will not allow log4j to open a file.

> But if I enable complex-thread behaviour the pgsrv crash at the
> second psql-session execution.
>
>   
Again, please send the backend log. If you don't want to disclose the 
content on this mailing list, send it to me directly.

Regards,
Thomas Hallgren