[Info-Ingres] [IJ]: DMF Stats Counter Reset

[Mark R. Winston]

Sounds like an Ingres Janitors project to me!  Karl, thanks for the
answer to the "how do I begin" question.  I've re-titled the thread and
included the "IJ" tag.

Paul (and any other interested parties), here's a link to "Thinking in
C" from Bruce Eckel <http://mindview.net/CDs/ThinkingInC/beta3>, and
here's a link to a text-based C tutorial
<http://www.linuxchix.org/content/courses/c_programming/>, both of which
I've found useful.

Mark R. Winston
www.datavailable.com



Betty & Karl Schendel wrote:
> At 2:37 AM +1100 11/16/07, Paul White wrote:
>> A new tracepoint.
>> I would consider it  a useful enhancement. It would make counting stats
>> a little easier without having to restart the server.
>> How much C do I have to learn?  Or where do I send the donuts and
>> coffee?
> 
> [re: resetting stats]
> 
> First you'd need to get a working build area;  that seems to range
> from trivially easy to wildly difficult, depending on your platform,
> your ability to follow directions, and a good helping of luck.
> 
> After that, you have to pick a tracepoint (probably a DMF tracepoint
> in this case, since you'll be resetting DMF stats, not all stats);
> do the right thing in dmf/hdr/dmftrace.h; add a dispatch to
> dmf/dmf/dmftrace.c;  and probably add the resetting routine to
> dmf/dmp/dm0p.c.  I say probably because I think that the buffer
> manager stats are semi-private, not sure that they are available
> in dmftrace.c.  If they are then you can ignore dm0p.c.
> The stats themselves are probably found in dmf/hdr/dm0pbmcb.h.
> This is all from memory so I might be wrong on some of it,
> can't be arsed to walk back to the "office" to make sure.
> 
> I don't know how to answer "how much C do I have to learn."
> You don't need to write lots of code from scratch, mostly
> just follow what's there.  One thing you WILL have to do is
> determine what mutexes (if any) need to be taken to access
> the various control blocks containing the stats, which
> probably means you'll have to at least skim some of dm0p.c
> to figure that out.
> 
> A trace point to reset ALL stats is impractical, since each
> facility owns its own.  I suppose you could invent an SCF
> tracepoint that calls each facility, but that's a BIG job;
> you would have to invent new "reset stats" entry points for
> all the backend facilities and that sounds like real work to me.
> 
> Karl
> _______________________________________________
> Info-Ingres mailing list
> Info-Ingres at kettleriverconsulting.com
> http://www.kettleriverconsulting.com/mailman/listinfo/info-ingres
>