Re: MYSQL Error 2013 load infile 15mln rec 6gb CSV

From: Mladen Gogala <gogala.mladen_at_gmail.com>
Date: Sun, 26 Oct 2014 16:56:40 +0000 (UTC)
Message-ID: <pan.2014.10.26.16.56.40_at_gmail.com>


On Sun, 26 Oct 2014 13:13:17 +0000, Eric wrote:

> On 2014-10-25, Mladen Gogala <gogala.mladen_at_gmail.com> wrote:
> 8>< Pretty much everything snipped...
>
> ... because responding to everything point by point is now on the verge
> of leading to incoherence.
>
> You said that the Postgres developers are idiots because they won't do
> hints. Then you suggested that it might be due to influence from a
> commercial fork.

I didn't call Postgres developers idiots, I called one particular Postgres developers an idiot. In my opinion, PostgreSQL developers would advance their careers by starting developing Drizzle or MariaDB, but that's another matter.

>
> In that case idiotic is probably not the right word. Did you make the
> suggestion in any of their forums?

The quoted article was written in response to the my arguments on the forum. I am the Oracle DBA mentioned in the article.

>
> I suspect that continuing that discussion would be going round in
circles.

I suppose so. As any DBA, coming from non-Postgres world can tell you, hints are a necessary evil. Oracle 12c has drastically reduced the need for hints, I haven't had the need to use them in 12c, but I suspect I will have to use them sooner or later. State of technology is just such that the users still have to have the manual override button. If PostgreSQL doesn't have it, than PostgreSQL should not be used in the enterprise environment. Please note that I am not talking about the commercial variants like EnterpriseDB, Vertica, Greenplum or Netezza, most of which have hints.

>
> However,
>

>> Nice try of changing subject.

>
> An attempt to draw a parallel - obviously wasted.

IMHO, there is no need for parallels, metaphors or any other novelistic elements. Parallels may produce a nice literary effect, but analogy is not considered logically correct method of deduction. The crux of our argument is whether there is a need for mechanism implementing manual override of the execution plans or not. After 25 years of being a DBA, I find the need for such mechanism self explanatory and natural. I have been in many situations in which hints have saved my skin. There is no need for parallels, metaphors or poetry here. If you can refute my argument and show me what do in a situation when a SQL suddenly stops performing and starts using wrong execution plan, I would be willing to accept your claims. Instead, I wee parallels and calls for "open mindedness". The first question I have to ask you is whether you are a DBA and have you ever been involved in an application development process of a large company?

>

>> The verb "to believe" means to hold some statements to be true, despite

>> of not having evidence to prove it.

>
> The verb has a number of dictionary definitions, you are taking an
> unduly narrow view.

verb (used without object), believed, believing. 1.
to have confidence in the truth, the existence, or the reliability of something, although without absolute proof that one is right in doing so: Only if one believes in something can one act purposefully.

http://dictionary.reference.com/browse/believe?s=t

>

>> Open mindedness isn't a virtue in itself,

>
> Yes it is.

Let's agree to disagree. Anyway, for this discussion I need proof that the manual override is not needed. Mind you, all other major relational databases implement such mechanism. PostgreSQL is the sole example of the database which does not allow that. That is why I have thrown it out of the server room of the company I used to work for. And that's about it.

-- 
Mladen Gogala
The Oracle Whisperer
http://mgogala.byethost5.com
Received on Sun Oct 26 2014 - 17:56:40 CET

Original text of this message