Message boards : Number crunching : many errors on my computer
Message board moderation
Author | Message |
---|---|
Send message Joined: 22 May 08 Posts: 49 Credit: 2,335,997 RAC: 0 |
My machine hostid=1170809 seems to error every HADAM3P model it tries, yet it processes the Coupled Model Full Resolution Ocean withput any problem. Is there some issue with the machine? |
Send message Joined: 17 Nov 07 Posts: 142 Credit: 4,271,370 RAC: 0 |
Hi NR, try setting some debug options in cc_config.xml . That might help solve the problem. Running boincmgr from a terminal window would allow you to see any boinc error messages. (Tip from GeoPhi.) Apart from that, try detaching from the project, deleting the HadAM3_xxx binaries, and re-attaching. HTH. |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
Dear Mr. Refractor, based on the error message, it looks like you're running too old a Linux distribution. See this series of posts from the beta site. Upgrade and prosper. |
Send message Joined: 7 Aug 04 Posts: 2187 Credit: 64,822,615 RAC: 5,275 |
There is a sticky in the Linux forum about this. Even though hadcm3n appears to complete successfully, the errors seen in stderr on the result page for those models are indicative of trouble zipping up the decadal upload files. I'm not sure that everything is getting up to the server, even though it says it completed successfully. hadam3p has immediate crashes on RedHat/CentOS 5 as you've seen. |
Send message Joined: 22 May 08 Posts: 49 Credit: 2,335,997 RAC: 0 |
Sorry for the late reply. Unfortunately I don't have admin rights on this machine. It's running Scientific Linux 5.4 (release date November 4, 2009)and is managed by the IT department. Does this mean that the machine is not producing useful scientific results for this project? If so, then unfortunately I will just have to detach the machine and attach it to some other project. |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
Sometime last year the department for which our 2 project people work, upgraded their Linux compiler. The new one uses a newer version of one of the library files than is used by some Linux distros, as per the sticky post to which George referred. So, no you're NOT producing useful results for the regional models, only for the coupled ocean models, which aren't being offered at the moment. Backups: Here |
Send message Joined: 22 May 08 Posts: 49 Credit: 2,335,997 RAC: 0 |
Sometime last year the department for which our 2 project people work, upgraded their Linux compiler. So the hadcm3n return valid data even with the errors in the stderr? |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
The one that I looked at said: Over Success Done so I didn't look at the error codes. This is getting further into the Linux version of the models than I understand, so I'll pass on a message and see if someone else knows. Backups: Here |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
It looks like RHEL, CentOS and Scientific are pretty conservative. So while they aren't old distributions, their kernels and associated libraries are already 18 to 24 months old at release. If your machine isn't running a too retrictive SELinux, you could try obtaining a copy of libstdc++.so.6 and see if it will run with BOINC after an LD_PRELOAD statement. Extract libstdc++.so.6.0.14 from this archive (click on the title link) and put it in a directory in your home folder, like /home/user-name/library-for-boinc. Rename it libstdc++.so.6 . Then in a terminal: LD_PRELOAD=/home/$USER/library-for-boinc/libstdc++.so.6 boinc . Edit: added rename instruction. |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
Okay, I've set up a Scientific 5.4 virtual machine here. I couldn't open the rpm that I linked (maybe has something to do with that compression issue geophi mentioned), so I used this slightly older version for CentOS, which extracted fine with the graphical tool. I was able to start BOINC with the LD_PRELOAD statement, but downloads have been hit and miss this morning, so I haven't finished downloading a hadam3p yet. Also my floating point and integer benchmarks are crazy, so I either need to install the VirtualBox additions, or the libstdc++.so.6 isn't going to work. I will post more information later. |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
Still waiting for this hadam3p to download due to the server issues. Meantime I fixed my benchmark results by adding divider=10 to to the guest's kernel parameters. |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
It works! Even under SELinux set to enforcing. The last rpm I linked did not work, so use this one instead (it will extract with the Scientific 5.4 gui tool.) To summarize: I installed BOINC by running the shell script for version 6.10.58 in my home directory. I think this is the way the OP is set up, since he's not an admin. I extracted libstdc++.so.6.0.10 from the rpm to my desktop. Then in a terminal (replace "user" with your user name): mkdir library-for-boinc cp Desktop/libstdc++.so.6.0.10 library-for-boinc/libstdc++.so.6 cd BOINC LD_PRELOAD=/home/user/library-for-boinc/libstdc++.so.6 nice -n 19 ./boinc & ./boincmgr & You can adjust the nice setting to your liking, or leave it out altogether. The terminal needs to be left running with BOINC set up this way. I think this can also work with BOINC installed as a service, but the boinc init script would need to have the LD_PRELOAD statement added to the line that starts boinc, and the libstdc++.so.6 has to go someplace where the boinc user can access it (/etc/boinc-client would be good). Edit: when starting out fresh on this version of Scientific Linux, you need this to get boincmgr working: su (enter root password) yum install libXcomposite |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
Got a trickle (and 100 posts.) |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
At the risk of turning this thread into a monologue, I'm reporting the successful completion of task 13967737. So in conclusion, one can run CPDN models on RHEL 5.1 to 5.5 (and its CentOS and Scientific derivatives) without changing the installed version libstdc++.so.6 (other applications will still use it happily.) |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
One more thing: contrary to my earlier instructions about leaving the terminal open, you can actually close it and even log out of your desktop session and boinc will run fine in the background. |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,039,635 RAC: 18,944 |
Don't think that last bit works if you are running BOINC from your somewhere in your ~ directory. |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
No illusions--it really worked. I had my BOINC directory in /home/user and closing the terminal seamlessly moved the parent process to 1. Note you have to start it with the ampersand at the end. I'll have try this on Ubuntu sometime soon. |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
I just tried it again and we're both right. You have to type 'exit' in the terminal, which is something I've gotten into the habit of doing for some reason. |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,039,635 RAC: 18,944 |
With Ubuntu - works after adding the ampersand at the end. Just means I don't get to look @ the errors in the terminal that I may at some point try and sort out. - They might hold clues as to why graphics don't work with BOINC7 for me. |
Send message Joined: 7 Aug 04 Posts: 2187 Credit: 64,822,615 RAC: 5,275 |
Thanks Belfry. That worked for me on my CentOS 5.7 PC. I did have to run "/bin/sh" first as it didn't recognize the LD_PRELOAD command from the terminal window with the default shell my user has (tcsh). Was the default shell for your user sh, or bash, or something else? |
©2024 cpdn.org