Re: Slow access to the data base.
From: Mark D Powell <Mark.Powell_at_eds.com>
Date: Mon, 19 Jan 2009 06:53:32 -0800 (PST)
Message-ID: <b937ca94-dfe4-4cd8-ba35-32d2472557bc_at_v5g2000prm.googlegroups.com>
On Jan 16, 2:30 pm, joel garry <joel-ga..._at_home.com> wrote:
> On Jan 16, 12:33 am, atch..._at_gmail.com wrote:
>
>
>
>
>
> > Good day,
>
> > I have an oracle installation on an IBM machine with AIX.
> > I have three users created on the Oracle DB: The user "oracle", "mo",
> > and "fo"
> > The installation was working well so far.
> > But for almost a week, the connexion of the users "mo" and "fo" have
> > been very slow. But the connexion to the user "oracle" has no
> > probleme.
> > When I stop the listener, I the users "mo" and "fo" can not access the
> > DB anymore wich mean that even on the same server, the users "mo" and
> > "fo" access the DB via a listenner.
>
> > This is my question:
>
> > 1- How can I do to make sure that on the server the users "mo" and
> > "fo" will access the DB without passing through a listener?
> > 2- What may have happened for the DB access to be sos slow?
> > 3- How can I globally solve the problem.
> > 4- Can I delete the users "mo" and "fo" and recreate them so that they
> > will connect to the DB without passing through a listener? How can I
> > make it?
>
> There are lots of things that can be happening here. metalink Note:
> 620256.995 has a person resolving identical symptoms by chmod 6751 on
> the oracle executable - it is a good example of what to check. If you
> don't have that exact protection on the oracle executable, it probably
> means you missed some installation steps, such as running root.sh. By
> the way, you don't want to set everything in the OH bin directory to
> that, just certain things.
>
> Off the top of my head (so season with a lot of salt), these are
> things you might check:
>
> Be sure you have all kernel patches and setting correct. It helps to
> say which exact version of Oracle and which exact version of the OS
> you are using whenever you post here (or anywhere, probably). Seehttp://dbaoracle.net/readme-cdos.htm#subj12 I have a vague memory
> that ibm has some docs online about certain AIX/oracle
> configurations. Your network hardware may even be relevant here.
>
> Check your tcp settings and also see what netstat has to say about who
> is connected. This is highly platform-dependent. Mark's assertion
> about using local listener connections and not having performance
> problems - well, to avoid the risk of spreading myths, I'm asking if
> anyone has any evidence one way or another. It's an assertion one of
> my vendors has made, and I don't think it is true over the universe of
> likely configurations and loads, but have no resources to prove it.
> And that vendor doesn't handle some tcp stuff correctly, I'm killing
> processes every day.
>
> Check your sqlnet.ora settings, some configurations may waste time.
> Also watch for platform-dependent bugs.
>
> AIX is somewhat idiosyncratic in my opinion, so take observations and
> advice about other platforms, even other unix and other versions of
> AIX, with caution. I don't have access to one so I can't be more
> specific, but when I did multi-platform support, hoo-boy.
>
> I'm sure I've seen more info out there on metalink and google for this
> sort of problem.
>
> jg
> --
> _at_home.com is bogus.
> If at first you don't succeed, throw good money after bad.http://www3.signonsandiego.com/stories/2009/jan/16/1m16ship001056-hop...- Hide quoted text -
>
> - Show quoted text -
Date: Mon, 19 Jan 2009 06:53:32 -0800 (PST)
Message-ID: <b937ca94-dfe4-4cd8-ba35-32d2472557bc_at_v5g2000prm.googlegroups.com>
On Jan 16, 2:30 pm, joel garry <joel-ga..._at_home.com> wrote:
> On Jan 16, 12:33 am, atch..._at_gmail.com wrote:
>
>
>
>
>
> > Good day,
>
> > I have an oracle installation on an IBM machine with AIX.
> > I have three users created on the Oracle DB: The user "oracle", "mo",
> > and "fo"
> > The installation was working well so far.
> > But for almost a week, the connexion of the users "mo" and "fo" have
> > been very slow. But the connexion to the user "oracle" has no
> > probleme.
> > When I stop the listener, I the users "mo" and "fo" can not access the
> > DB anymore wich mean that even on the same server, the users "mo" and
> > "fo" access the DB via a listenner.
>
> > This is my question:
>
> > 1- How can I do to make sure that on the server the users "mo" and
> > "fo" will access the DB without passing through a listener?
> > 2- What may have happened for the DB access to be sos slow?
> > 3- How can I globally solve the problem.
> > 4- Can I delete the users "mo" and "fo" and recreate them so that they
> > will connect to the DB without passing through a listener? How can I
> > make it?
>
> There are lots of things that can be happening here. metalink Note:
> 620256.995 has a person resolving identical symptoms by chmod 6751 on
> the oracle executable - it is a good example of what to check. If you
> don't have that exact protection on the oracle executable, it probably
> means you missed some installation steps, such as running root.sh. By
> the way, you don't want to set everything in the OH bin directory to
> that, just certain things.
>
> Off the top of my head (so season with a lot of salt), these are
> things you might check:
>
> Be sure you have all kernel patches and setting correct. It helps to
> say which exact version of Oracle and which exact version of the OS
> you are using whenever you post here (or anywhere, probably). Seehttp://dbaoracle.net/readme-cdos.htm#subj12 I have a vague memory
> that ibm has some docs online about certain AIX/oracle
> configurations. Your network hardware may even be relevant here.
>
> Check your tcp settings and also see what netstat has to say about who
> is connected. This is highly platform-dependent. Mark's assertion
> about using local listener connections and not having performance
> problems - well, to avoid the risk of spreading myths, I'm asking if
> anyone has any evidence one way or another. It's an assertion one of
> my vendors has made, and I don't think it is true over the universe of
> likely configurations and loads, but have no resources to prove it.
> And that vendor doesn't handle some tcp stuff correctly, I'm killing
> processes every day.
>
> Check your sqlnet.ora settings, some configurations may waste time.
> Also watch for platform-dependent bugs.
>
> AIX is somewhat idiosyncratic in my opinion, so take observations and
> advice about other platforms, even other unix and other versions of
> AIX, with caution. I don't have access to one so I can't be more
> specific, but when I did multi-platform support, hoo-boy.
>
> I'm sure I've seen more info out there on metalink and google for this
> sort of problem.
>
> jg
> --
> _at_home.com is bogus.
> If at first you don't succeed, throw good money after bad.http://www3.signonsandiego.com/stories/2009/jan/16/1m16ship001056-hop...- Hide quoted text -
>
> - Show quoted text -
Evidence is available via the sqlnet trace and listener.log file. If the connects are truely slow then it becomes an issue of finding the mis-configuration that will solve the problem. Your note is a good starting point.
HTH -- Mark D Powell -- Received on Mon Jan 19 2009 - 08:53:32 CST