Hi Andrej,

On Mon, 15 Oct 2001, Andrej Filipcic wrote:

> On Tue, 16 Oct 2001, Matt Sullivan wrote:
> 
> I have the same problem with Waiting for resolver in threaded mode. Over
> the weekend I have tried what you suggested, but the problem still
> persists. I have noticed that if there is a request which returns "no
> servers could be reached", the zombie appears. It helps if I increase the
> timeout from 10 to 100 seconds in GetHostByName call in filters.cpp.
> However, index -E or kill still has problems...

Can you tell me what you have configured in your /etc/resolv.conf on your
indexing box?  It sounds from what you've said as though your nameservers are a
bit flakey.
 
I would highly recommend running a nameserver on your indexing box which
answers queries on localhost (127.0.0.1:53) - ultimately you may configure it
as a forwarder only but at least index will not ever be in a situation where
your nameserver is unreachable. 

If possible, could you give it a try and let me know how it goes. 


Matt.


> The asp rpm package has the same problems (rh 7.1)
> 
> Best regards,
> 
> Andrej
> 
> > Hi,
> >
> > Could anyone who is currently having issues that appear to be resolver related
> > please try the following and let me know if this fixes the problem.
> >
> > * cd into your build directory and perform a "make distclean".
> >
> > * edit the file "include/aspseek-cfg.h.in" and delete the two lines that read:
> >
> >   /* Define if you have gethostbyname_r function */
> >   #undef HAVE_GETHOSTBYNAME_R
> >
> > * edit the file "src/resolve.cpp" and delete the two lines that read:
> >
> >   // Broken my_gethostbyname_r
> >   #undef HAVE_GETHOSTBYNAME_R
> >
> > * configure the package as you would normally i.e. run ./configure etc. and
> >   then "make" and or "make install" as usual.
> >
> > Finally, run the indexer and test for continuation of resolver issues.
> >
> > Please let me know if this corrects the issues that you are having.
> >
> >
> > Thanks,
> > Matt.
> >
> > On Fri, 5 Oct 2001, Elsner wrote:
> >
> > > At 18:26 05.10.2001, you wrote:
> > >
> > > >5 Cctober 2001 19:07,  You wrote:
> > > > > As there was a thread dealing with this message, I am facing the same
> > > > > problem:
> > > > >
> > > > > Start index with:
> > > > > sbin/index -N 15 -R 7 -m
> > > > > (seems reasonable, as number of resolvers is half the number of
> > > > > indexers)
> > > > >
> > > > > After a while, 'Waiting for resolver' messages flood the screen.
> > > > > index -E will not stop the indexers, used a 'killall index' then.
> > > > > And then 'killall -KILL index' as last resort.
> > > >
> > > >
> > > >Are these two problems connected together??
> > > >I have unstoppable threads even if  I  run more resolvers and never get
> > > >"waiting for resolver" message.
> > > >
> > > >Frank, do you also have zombies in your system while running multithreaded
> > > >index as I do ?
> > >
> > > After issueing 'kill KILL index' nothing remains in the process table,
> > > use 'ps -aef | more' to scan through. What do you mean by 'zombies',
> > > these should be marked by '<defunct>''? I do not see any of them.
> > >
> > > I will have another try on Monday. It is late in Germany now for work ...
> > > (18.30 Local Time).
> > >
> > > Frank
> > > #-------------------------------------------------------#
> > > Dipl.-Math. Frank Elsner
> > > Universitaet Osnabrueck (University of Osnabrueck)
> > > - Rechenzentrum - (Computing Center)
> > > Albrechstrasse 28, AVZ
> > > D-49076 Osnabrueck
> > > Deutschland (Germany)
> > >
> > > Tel. (Phone): ++49 (0)541/969-2343 Fax: -2470
> > > E-Mail: [EMAIL PROTECTED]
> > > #-------------------------------------------------------#
> > >
> >
> 
> -- 
> _____________________________________________________________
>    doc. dr. Andrej Filipcic,   E-mail: [EMAIL PROTECTED]
>    Department of Experimental High Energy Physics - F9
>    Jozef Stefan Institute, Jamova 39, P.o.Box 3000
>    SI-1001 Ljubljana, Slovenia
>    Tel.: +386-1-477-3674    Fax: +386-1-425-7074
> -------------------------------------------------------------
> 

Reply via email to