Re: auto_explain logs ERROR: 08P01 for every query using bind variable

On Fri, Jan 31, 2020 at 2:14 AM Ted Liu <ted_liu(at)yahoo(dot)com> wrote:

> To whom it may concern,
>
> For performance and security, most of our queries use bind variables. We
> also implemented auto_explain for tracking down un-optimized queries. Our
> database supports a large application. During the work-hours, there are
> 30,000+ queries executed every hour. PostgreSQL logs rotates every hour and
> each log contains 30,000+ error message like below.
>
> 2020-01-27 00:00:03 UTC:xxx.xx.0.189(35556):xxxxxx(at)xxxxxxdb:[130756]:ERROR:
> ERROR: 08P01: bind message supplies 0 parameters, but prepared statement
> "..." requires 1
> 2020-01-27 00:00:03 UTC:xxx.xx.0.189(35558):xxxxxx(at)xxxxxxdb:[130756]:LOCATION:
> exec_bind_message, postgres.c:1601
> 2020-01-27 00:00:03 UTC:xxx.xx.0.189(35558):xxxxxx(at)xxxxxxdb:[130756]:STATEMENT:
> EXPLAIN select ...,...,..,... from ... where ....=$1
>
> These error messages are really overwhelming to us and they are not
> useful at all. In our opinion, these error messages should not be logged.
>

I agree they shouldn't be logged. And I don't get such messages when using
auto_explain with something like `pgbench -M prepared`.

What version are you using? What settings are you using for auto_explain?
What is your client language/driver? Maybe the client is doing something
weird like sending duplicate binds?

Cheers,

Jeff