From owner-chemistry@ccl.net Wed Oct 28 13:36:01 2009 From: "Denis Gulin deninho++quimica.ufpr.br" To: CCL Subject: CCL:G: Running gaussian on a cluster ... Message-Id: <-40544-091028133310-23622-f41WYhYubFOFcbe0tZsNbw{:}server.ccl.net> X-Original-From: "Denis Gulin" Content-Transfer-Encoding: 8bit Content-Type: text/plain;charset=iso-8859-1 Date: Wed, 28 Oct 2009 12:21:26 -0200 (BRST) MIME-Version: 1.0 Sent to CCL by: "Denis Gulin" [deninho-.-quimica.ufpr.br] Hi CCLers, Recently we've upgrade Gaussian03 to Revision E.01. The compilation (with Portland compilers) has finished fine. Jobs also run fine in the main node of our cluster but, when we try run Gaussian on other node (processor servers), it fails. These other nodes doesn't have local hard disks so, the system is mounted from the main node, as NFS. When the jobs fail, we receive the error massage: "buffer allocation failed in ntrext1.: Cannot allocate memory" and a line "buffer allocation failed in ntrext1." is printed in the log file. My question is: There is something to do, for this 'limitation' for running Gaussian on networking file systems, like NFS ? How can I overcome this problem ? Any ideas ? Thanks in advance, Regards, -- Denis Jeison Gulin -- deninho-*-quimica.ufpr.br Laboratório de Química Computational Universidade Federal do Paraná - UFPR Centro Politécnico P.O.Box 19081 81531-990 Curitiba/PR - Brazil From owner-chemistry@ccl.net Wed Oct 28 15:16:00 2009 From: "Eduardo Lemos de Sa edulsa*quimica.ufpr.br" To: CCL Subject: CCL:G: Problems to run G03 in a cluster Message-Id: <-40545-091028125926-19801-DZuXyjDIq5ZRCSkF9O66hA|,|server.ccl.net> X-Original-From: "Eduardo Lemos de Sa" Content-Transfer-Encoding: 8bit Content-Type: text/plain;charset=iso-8859-1 Date: Wed, 28 Oct 2009 12:15:56 -0200 (BRST) MIME-Version: 1.0 Sent to CCL by: "Eduardo Lemos de Sa" [edulsa[a]quimica.ufpr.br] Hi, Recently we've upgrade Gaussian03 to Revision E.01. The compilation (with Portland compilers) has finished fine. Jobs also run fine in the main node of our cluster but, when we try run Gaussian on other node (processor server), it fails. These other nodes doesn't have local hard disks so, the system is mounted from the main node, as NFS. When the jobs fail, we receive the error massage: "buffer allocation failed in ntrext1.: Cannot allocate memory" and a line "buffer allocation failed in ntrext1." is printed in the log file. My question is: There is something to do, for this 'limitation' for running Gaussian on networking file systems, like NFS ? How can I overcome this problem ? Any ideas ? Thanks in advance, Regards, Eduardo -- Eduardo Lemos de Sa edulsa%x%quimica.ufpr.br Professor Assoc. II - Departamento de Quimica Universidade Federal do Parana C. P. 19081 81531-990 Curitiba/PR - Brazil Fone (41) 3361 3300 From owner-chemistry@ccl.net Wed Oct 28 16:58:01 2009 From: "Lu Wu wlu3*uky.edu" To: CCL Subject: CCL:G: about the memory size used in Gaussian03 Message-Id: <-40546-091028142607-16485-7zxWW4preL+cmKa36ykqBw!^!server.ccl.net> X-Original-From: "Lu Wu" Date: Wed, 28 Oct 2009 14:26:01 -0400 Sent to CCL by: "Lu Wu" [wlu3(!)uky.edu] Hi- Recently I performed the calculation of PrO using b3lyp method with ANO-RCC basis set, and used 32GB of memory with 16processes. But the job always stopped with the error message: not enough memory. Could you tell me how to estimate the memory size we need? the error message is: ------------------- The (ff|fd) Classes Not enough memory for any path: MxKBra= 121 MxKKet= 165 Memory= 265504377. After MaxDer, Avail=FFFFFFFFFFFFFFFFFFFFFFFFFFFFFTFFFFFFFFFFFFFFFFFFFF After MemPth, Avail=FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF After MxKxxx, Avail=FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF After UseAlp/DoOS, Avail=FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF No path selected in FillAv. Error termination via Lnk1e in /share/cluster/SLES9/ppc64/apps/gaussian/E.01/g03/l703.exe at Fri Oct 23 22:38:51 2009. Job cpu time: 1 days 7 hours 46 minutes 32.1 seconds. File lengths (MBytes): RWF= 158 Int= 0 D2E= 0 Chk= 3 Scr= 4 Thank you Lu From owner-chemistry@ccl.net Wed Oct 28 23:01:01 2009 From: "Zhong jie liang zjliang(!)mail.shcnc.ac.cn" To: CCL Subject: CCL: promble about ONIOM Message-Id: <-40547-091028225024-25110-jG+2aJsEGotFu9VG/hR7PA|a|server.ccl.net> X-Original-From: "Zhong jie liang" Date: Wed, 28 Oct 2009 22:50:20 -0400 Sent to CCL by: "Zhong jie liang" [zjliang(_)mail.shcnc.ac.cn] Dear experts: I have had the problem for a long time. I have been doing the potential energy surface scan on a deacetylase and been successful on it in the bacteria, But when I used the same method on the protein in yeast, I cant get a reasonable result. I have changed the scan range and changed the QM atoms, but the potential energy surface was always ascending, without any down trend and any transition state point. At last, I did optimization of the product of the first reaction I need. Surprisingly, it was optimized for a long time and went back to the reactant. I wondered if it was the problem on charges. Now, I attached the product optimization input file and the scan input file in this letter. Wish you could help me . Any suggestion from you would be appreciated very much. And I wonder are there any other key words useful in TS found such as QST2, QST3 and how to use them. The product optimization input file as follows: %chk=p.chk %mem=2000Mb %nproc=4 # opt=(gdiis,loose,micro) oniom(b3lyp/6-31g(d):amber=hardfirst) geom=connectivity p -4 1 1 1 1 1 N-N3-0.159200 0 11.496194 -16.994614 -0.728200 L C-CT-0.022100 0 10.825913 -17.794455 -1.776714 L C-C-0.612300 0 10.312565 -16.884910 -2.894192 L O-O--0.571300 0 9.767045 -15.814949 -2.636844 L C-CT-0.086500 0 9.705633 -18.671998 -1.195298 L C-CT-0.033400 0 9.022321 -19.520116 -2.275051 L