Hey, thanks heaps for the information, I didn't know about "alter system
register" yet so looked it up and found the answer to the timing of the pmon
registration which is 1 minute (previously I had found this out from
metalink article 69546.1)
REGISTER Clause
Specify REGISTER to instruct the PMON background process to register the
instance with the listeners immediately. If you do not specify this clause,
then registration of the instance does not occur until the next time PMON
executes the discovery routine. As a result, clients may not be able to
access the services for as long as 60 seconds after the listener is started.
In terms of static registration, I do have this set and so the website is
always available, however, for intermedia/text searching, I have to use
dynamic registration (until I can get it working with static registration)
as I can't get 8.1.6 external procedure calls to work with static
connections. This has something to do with running a 9i listener (I think)
which is possibly calling the 9i extproc binary instead of the 8.1.6 one so
I get the following errors:
SQL> exec ctx_output.start_log('log');
BEGIN ctx_output.start_log('log'); END;
*
ERROR at line 1:
ORA-20000: interMedia Text error:
ORA-28576: lost RPC connection to external procedure agent
ORA-06512: at "CTXSYS.DRUE", line 126
ORA-06512: at "CTXSYS.CTX_OUTPUT", line 39
ORA-06512: at line 1
I'm still looking into this.
Anyway, thanks for the helpful information.
"Howard J. Rogers" <howardjr2000_at_yahoo.com.au> wrote in message
news:gq2ca.2518$LT.6674_at_newsfeeds.bigpond.com...
> "Peter" <no_email_at_no_email.com> wrote in message
> news:b4p9hu$juf$1_at_bunyip.cc.uq.edu.au...
> > The documents that I quoted, specifically state that, "without actually
> > stopping the listener" which is different to what you said which is,
> > "Basically a reload is a stop followed by a start".
> >
> > I wish that the documents were more specific in that they don't state
(or
> I
> > can't find) when the dynamic services will be registered after the
reload
> > command is issued
>
> Check the documentation for the PMON process, because it's PMON that
> actually does the instance registration on a periodic basis (last time I
> checked, it was about every 5 minutes).
>
> >- "...listener will be unregistered and subsequently
> > registered again...". I had assumed that this would be straight away but
I
> > have found that this is not the case.
> >
> > With the ever increasing use of dynamic registration, I believe that the
> > documentation is lacking, and could cause downtime, in that it should
> state
> > that dynamically reliant services could be unavailable for up to a
minute
> at
> > a time (and I have found this to be longer when the listener is
> experiencing
> > high loads). Instead this is not mentioned at all.
> >
> > This brings me back to my original question, which is, does the lsnrctl
> > reload command drop dynamic service handlers thus making dynamically
> > registered databases unavailable for a certain period of time (a
> supposedly
> > maximum period of 1 minute)?
>
> Er, sort of. Bear in mind that you can have dynamic registration happening
> (for the purposes of load-balancing) and yet still have databases
statically
> declared (still required for Enterprise Manager, for example). So you
might
> lose load-balancing capabilities for potentially up to 5 minutes (not sure
> where you're getting the 1 minute figure from), but connections would
still
> be available during that time because of the static declarations.
>
> Or you could migrate to 9i, where the command 'alter system register'
forces
> re-registration at will.
>
> (Incidentally, in case there was any doubt, the stopping and starting of a
> listener never means that existing connections are killed off, merely that
> new ones cannot be made for the duration. So the statement "This would of
> course result in wiping mts connections." would be wrong if "wiping" were
to
> be taken to mean 'terminating").
>
> Regards
> HJR
>
>
>
>
Received on Thu Mar 13 2003 - 18:29:28 CST
