IDS Forum

Re: onstat -l

Posted By: Fernando Nunes
Date: Wednesday, 8 December 2010, at 6:19 p.m.

In Response To: Re: onstat -l (Art Kagel)

Nice... I started this reply by saying I haven't seen this output either...
But then I went into the knowledge base...
You may not believe it, but it's there since version 4.x at least. (tbstat
-l) :)
It was about time that someone documents it (assuming it isn't - I didn't
look - )

L-x means the logical buffer... Please confirm if you ever seen L-4, since
Art is right, it should be at most L-3
ioproc should be the thread identifier... No idea what "0" means. No idea
what the flags mean...

Do you have HDR or SDS setup? When you see this output you should also have
"G" flags on the onstat -u output.
This can be a serious problem if you have it frequently.
I believe there is material around this issues (when they are issues, not
saying that's your case) to write a small book :)

In HDR environments this is one of the reasons why the primary can be
impacted by secondary. We've seen several customer having problems, but too
many times the cause is related to poor configuration or too much load on
the secondary (there was a chat with the labs session that dig into this).
So, I think there hasn't been enough "pressure" to solve this, and I suppose
there is no decision about the solution. Typically this is seen in
unbuffered logging databases together with high usage systems.
In unbuffered logging, each commit forces a logical log buffer flush, and
with many sessions this sometimes makes the 3 logical log buffers not enough
for the rotation needed. A simple solution would be to get more logical log
buffers, but I believe R&D would prefer a better solution.

Not much help, but if you require documentation on this, please open a PMR
and ask about it (I know customers have other/better things to do, but
that's the quickest way to get it done).

Regards.

On Wed, Dec 8, 2010 at 1:59 AM, Art Kagel <art.kagel@gmail.com> wrote:

> There are three logical log buffers: L-1, L-2, & L-3. The current buffer
> rotates among these three. While one is being written to disk the next one
> is used, etc. I don't remember seeing this output, but it may indicate that
> one or both of the buffers which is not current are waiting to be written
> out. This can become a problem if the IO subsystem is so slow that all
> three buffers are full and logical log records must wait to be written to a
> buffer until one can be flushed and released for reuse. What are the IO
> service times reported for the logical log chunk(s) in onstat -g iof? If
> these are high, say over 0.025secs, you may have an IO bottleneck on those
> chunks.
>
> Art
>
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> IIUG Board of Directors (art@iiug.org)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinions
> and
> do not reflect on my employer, Advanced DataTools, the IIUG, nor any other
> organization with which I am associated either explicitly, implicitly, or
> by
> inference. Neither do those opinions reflect those of other individuals
> affiliated with any entity with which I am affiliated nor those of the
> entities themselves.
>
> On Tue, Dec 7, 2010 at 6:58 PM, Tristan Ball <tristanb@pronto.com.au>
> wrote:
>
> > I'm seeing a couple of extra lines in the output of 'onstat -l', which I
> > can't
> > find documented in the 11.50 administrators guide and references.
> >
> > The lines in question look like this:
> >
> > Buffer Waiting
> > Buffer ioproc flags
> > L-1 0 0x1 0
> > L-2 0 0x1 0
> >
> > The exact lines varies (L-1 through to L-4 I think), and sometimes the
> > entire
> > block simply isn't there.
> >
> > My naive assumption is that this probably represents a problem, and if
> > something is waiting on buffers long enough for it to show in onstat -l
> for
> > several seconds at a time?
> >
> > Is anyone familiar with what this actually means?
> >
> > Thanks,
> >
> > Tristan
> >
> >
> >
> >
>
> *******************************************************************************
> > Forum Note: Use "Reply" to post a response in the discussion forum.
> >
> >
>
> --485b393ab8495e10a10496dc775c
>
>
>
> *******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>

--
Fernando Nunes
Portugal

http://informix-technology.blogspot.com
My email works... but I don't check it frequently...

--0015174c1102864fb90496ee5968

Messages In This Thread

• onstat -l
  Tristan Ball -- Tuesday, 7 December 2010, at 6:58 p.m.
  • Re: onstat -l
    Art Kagel -- Tuesday, 7 December 2010, at 8:59 p.m.
    • RE: onstat -l
      Tristan Ball -- Wednesday, 8 December 2010, at 4:41 p.m.
      • Re: onstat -l
        Art Kagel -- Wednesday, 8 December 2010, at 6:05 p.m.
        • RE: onstat -l
          Tristan Ball -- Wednesday, 8 December 2010, at 8:25 p.m.
          • Re: RE: onstat -l
            Art Kagel -- Wednesday, 8 December 2010, at 9:01 p.m.
    • Re: onstat -l
      Fernando Nunes -- Wednesday, 8 December 2010, at 6:19 p.m.
      • RE: onstat -l
        Tristan Ball -- Wednesday, 8 December 2010, at 9:07 p.m.
        • Re: onstat -l
          Fernando Nunes -- Wednesday, 8 December 2010, at 9:44 p.m.

IDS Forum is maintained by Administrator with WebBBS 5.12.