sum: setup cluster & avoid g98 error l9999



Dear CCLers:
      Some time ago I posted the question about how to setup a cluster and
 how to avoid the error of gaussian 98 link 9999. Now I have solved the
 error l9999 problem by opt(calcall). Thanks for ccl and you all.
 I wrote the following question:
 ---------- Forwarded message ----------
 From: "Fan Jianmiao" <jim_fan;at;263.net>
 To: <CHEMISTRY;at;ccl.net>
 Sent: Thursday, March 22, 2001 8:16 PM
 Subject: CCL:setup a cluster and avoid g98 error link 9999
 > Dear CCLers:
 >
 >      I want to get some help on how to establish a cluster and how to avoid
 > the error of gaussian 98 link 9999.
 >
 >      Our lab needs a cluster running gaussian to get a better calculating
 > envirement. But we haven't enough money, we want to buy some PC and form a
 > small cluster. But now I don't know the price of Gaussian in parallel(can
 it
 > sell to china?) and whether there is any other programs needed. Currently
 we
 > have buy four PCs(each is equipted with 2 processors of pIII933 and 768M
 > SDRAM) and a network switcher. But I haven't installed any software for
 > paralleling yet. I wonder, how shall I make up the system?
 >
 >      Another problem is how to avoid the error of g98 link 9999. This error
 > is very common in our work. Usually we change the iop parameter, but it
 works
 > not alwaysly. We found if the start geometry is not very good then maybe
 this
 > problem will be conquered. And if the start geometry is similler to the
 finial
 > , it will not come to a normail termination then. Why?  Puzzling problem!
 > I often saw this infomation on my screen:
 > " Error termination request processed by link 9999.
 >   Error termination via Lnk1e in /usr/g98/l9999.exe.
 >   Job cpu time:  1 days  1 hours 39 minutes  8.0 seconds.
 >   File lengths (MBytes):  RWF=  824 Int=    0 D2E=    0 Chk=    7 Scr=
 1"
 >
 > What shall I do after I failed changing the iop parameter?If the geometry
 should
 > not be wrong, how can I manage to overcome it?
 >
 >     If there's anyone can give me some suggestion or comments on it, I
 would
 > be very appreciated.
 >
 >     Thanks for all in advance.
 > With My Best Regards,
 >
 > Sincerely yours,
 > Jim Fan
 I received the following answers
 ---------- Forwarded message ----------
 Date: Thu, 22 Mar 2001 23:06:43 +0100
 From: Bernd Schubert <bernd.schubert;at;student.uni-halle.de>
 To: Fan Jianmiao <jim_fan;at;263.net>
 Subject: Re: CCL:setup a cluster and avoid g98 error link 9999
 Hi,
 I think for running gaussian parallel on a cluster you need the LINDA program
 sold by the gaussian people (for pricing have a look at:
 http://www.gaussian.com/linda.htm)
 I think your problem with link 9999 depends on the type of job you run. I'm
 doing a lot of dft calculations and very often the energy of of the geometry
 optmisations is not converging. On the contrary the force constants and the
 energy begin to oscillate. Without a maxiumum of geometry optimisations such
 job could run for a very long time without finding an end.
 If you are also doing this kind of jobs (dft or hf calculations) then have a
 look on your output file. At the end of the last geometry should be mentioned
 that the maximum number of geometry optimisations is exceeded.
 The maximum number of geom. calculations is controlled by the keyword
 opt(MaxCyc=N).
 To make the energy better converging it is very often helpful to calculate
 the hesse matrix controlled by the keyword opt(calcfc) - only before the
 first geom. optmisation or by opt(calcall) - before every geom. optimisation
 - or by iop(1/40=N) - every N steps, also possible when analytical
 calculations of the hesse matrix are not possible.
 An other (and additional to the first suggestions) possibility to make
 gaussian better finding the minimum is to reduce the maximum stepsize between
 the geometry optimisations controlled by iop(1/8=N) - opt(stepsize=N) is not
 working. I have made good experience with iop(1/8=10).
 I hope that this is your problem and that it might help you,
 Best regards,
 Bernd
 ---------- Forwarded message ----------
 Date: Thu, 22 Mar 2001 22:28:55 -0800
 From: Roy Jensen <royj;at;uvic.ca>
 To: Jianmiao Fan <fan;at;mail.ustc.edu.cn>
 Subject: Re: How to avoid g98 error link 9999
 On Fri, 23 Mar 2001 13:05:25 +0800 (CST), you wrote:
 > Dear Jensen:
 >    Thanks for your help.I have attched the input and output file.
 > The job is terminated by steps exceeded. How can I get a normal
 > termination?
 >
 > On Thu, 22 Mar 2001, Roy Jensen wrote:
 >
 > > Jim
 > >
 > > What is the input line for the failed job? Send me that and the
 > > complete output file and I will see what I can do about your error
 > > termination problem.
 > >
 > > Roy Jensen
 > >
 >
 > With My Best Regards,
 >
 > Sincerely yours,
 >          Jim Fan
 Jim
 The input geometry appears to be too far from the optimized structure.
 Gaussian calculates the force constants once and uses them for the
 remainder of the run. You have two options:
 1) Step 19, 22, and 33 came close to converging. Use any of these
 geometries as the initial geometry and restart the job.
 2) Use opt(calcall) to get gaussian to calculate the force constants
 at ever step. You do not need to do a second freq calculation; it is
 automatically done with calcall. This method is time consuming but
 should work for most any input geometry.
 Best of luck,
 Roy Jensen
 ---------- Forwarded message ----------
 Date: Wed, 28 Mar 2001 09:42:52 -0500
 From: Jerry Lipchus <lipchus;at;sca.com>
 To: Fan Jianmiao <jim_fan;at;263.net>
 Subject: Linda for Gaussian
 Dear Fan Jianmiao(Jim_Fan) ,
 Linda is available for download at :
 http://www.sca.com/site/downloads/index.html
 LINDA License
 xxxx xxxx xxxx xxxxxx xxxxx xxxxxx x xxxx xxxx
 Midnight of May 28, 2001, exclusive
 --
 Scientific Computing Associates
 265 Church Street
 New Haven, CT 06511
 Technical Support       lsupport;at;sca.com
 Tel: (203)777-7442      Fax: (203)776-4074
 Thank You
 Best regards,
 Jerry Lipchus
 Director of Sales
 Scientific Computing Associates, Inc.
 265 Church Street
 Suite 1010
 New Haven, CT 06510
 lipchus;at;sca.com
 203-777-7442 ext 275
 ------------------------------------------------
 ------------------------------------------------
 With My Best Regards,
 Sincerely yours,
 Jim Fan
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 Fan Jianmiao(Jim_Fan)
 Department of Chemistry
 East Campus,USTC
 Hefei,Anhui,230026,P.R.China
 Tel:86-551-3654322(dorm) E-Mail:jim_fan;at;263.net
     86-551-3606640(lab)  Http://boc.ustc.edu.cn/members/jim_fan/
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 I am an undergraduate student.