Smp

[Adak]

No rules covered that kind of problem. As long as it affected all the teams, I guess no rule would be needed.

When FAH goes on the fritz, I generally shut down the systems for maintenance, shut 'em down and enjoy the silence, or crunch a little SETI.

I think it's pure crap what happened to you with all those WU's. Stanford should have a backup for each one of their servers, all set to go with redundant data drives. You can't expect a server to run 24/7 and never have to shut down.

Hundreds of thousands of systems sending in results every week or two, and they can't keep a backup server ready to handle the load? Not good.

[ThunderRd]

Well, it's 10AM here and still the server is down status. Meh.

I think I'll just shut down and wait it out. There's no reason to keep working til they get it sorted out. The WUs just get stale and expire.

[dabaerman]

so far so good. installed and running on one machine. on to the others. question's so far, will the client run only in systray? can it be hidden?

edit:machine 2 has mpiexec error when I try to run fah.exe. anything to do with 'account name' ? I left that blank and without a password.
going for #3.

Ron

[ThunderRd]

Originally Posted by dabaerman

so far so good. installed and running on one machine. on to the others. question's so far, will the client run only in systray? can it be hidden?

Ron

It doesn't run in systray- perhaps you mean on the taskbar?

If that is what you are talking about, then the answer is no, it can't be hidden *unless* you run it as a service, in which case I would suggest that you observe the client and make sure it's ok for awhile before trying that. It's a slightly complicated procedure involving making changes in the service properties dialogs, and it is an unsupported thing.

It also would mean that you couldn't see what the client is doing, which is why I'd suggest practicing with it in a window first, so you are confortable with it.

[ThunderRd]

@Ron: you must install under your windows admin credentials; that means that you MUST have a password to log into windows. If you don't have a password, go to user accounts and create one for the admin account. Then, uninstall everything you have done and start again, following the procedure I gave you again.

When the client configuration asks you for a credential store password, make it blank. But when MPI later asks you to enter your username/password twice, hit return for username, and then enter your logon password.

There is no way around this. The client WILL NOT RUN without a windows logon password.

[ThunderRd]

And now...

OK, at around noon here I saw that the servers were up again, and I held back on shutting the farm down for awhile to see if I can get some work uploaded in time, but nothing is happening. I'm stumped, because before these last 2 days all the machines in question were happily down/uploading without problems.

And now I see that I'm having the problem on .65.64:8080 as well as .65.63:8080. These are the addresses of 2 of the work servers. We download on port 80, which is operating fine for me, but we upload to them on port 8080, which apparently is my problem. They also have collection servers, which also use port 8080. Can't get to them either.

I have done nothing to the configuration of the router, and I am about to try to call my ISP to find out if port 8080 is closed for some reason. It's the only reason I can think of that would suddenly prohibit 20 boxes from communicating to the servers. I can ping the servers, and I can download from them fine on :80. But nothing is happening in the other direction, and I can't contact the server at :8080 in my browsers.

Has anybody ever had an ISP close port 8080, and why would they? If they have closed it, then it's the coincidence of the month that it would happen just as the servers went down - confusing me even further. I'm afraid to call and ask 'cause I'll most likely get some PFY who will spin my wheels and just get me aggravated.

--- Opening Log file [May 21 06:55:12] 


# SMP Client ##################################################################
###############################################################################

                       Folding@Home Client Version 5.92beta

                          http://folding.stanford.edu

###############################################################################
###############################################################################

Launch directory: D:\f@h_smp30
Executable: D:\f@h_smp30\fah.exe
Arguments: -verbosity 9 

[06:55:12] - Ask before connecting: No
[06:55:12] - User name: ThunderRd (Team 45)
[06:55:12] - User ID: 4DFB3331571CB210
[06:55:12] - Machine ID: 1
[06:55:12] 
[06:55:12] Loaded queue successfully.
[06:55:12] 
[06:55:12] - Autosending finished units...
[06:55:12] + Processing work unit
[06:55:12] Trying to send all finished work units
[06:55:12] Core required: FahCore_a1.exe


[06:55:12] Core found.
[06:55:12] + Attempting to send results
[06:55:12] - Reading file work/wuresults_05.dat from core
[06:55:12] Working on Unit 06 [May 21 06:55:12]
[06:55:12] + Working ...
[06:55:12]   (Read 5525230 bytes from disk)
[06:55:12] - Calling 'mpiexec -channel shm -env MPICH_USE_SMP_OPTIMIZATIONS 1 -np 4 FahCore_a1.exe -dir work/ -suffix 06 -checkpoint 15 -verbose -lifeline 2316 -version 592'

[06:55:12] Connecting to http://171.64.65.64:8080/
[06:55:14] 
[06:55:14] *------------------------------*
[06:55:14] Folding@Home Gromacs SMP Core
[06:55:14] Version 1.76 (February 23, 2008)
[06:55:14] 
[06:55:14] Preparing to commence simulation
[06:55:14] - Ensuring status. Please wait.
[06:55:31] - Looking at optimizations...
[06:55:31] - Working with standard loops on this execution.
[06:55:31] Examination of work files indicates 8 consecutive improper terminations of core.
[06:55:31] es status OK
[06:55:33] - Couldn't send HTTP request to server
[06:55:33] + Could not connect to Work Server (results)
[06:55:33]     (171.64.65.64:8080)
[06:55:33] - Error: Could not transmit unit 05 (completed May 20) to work server.
[06:55:33] - 8 failed uploads of this unit.


[06:55:33] + Attempting to send results
[06:55:33] - Reading file work/wuresults_05.dat from core
[06:55:33]   (Read 5525230 bytes from disk)
[06:55:33] Connecting to http://171.64.122.86:8080/
[06:55:40] 39165 -> 12883625 (decompressed 528.1 percent)
[06:55:40] 8.1 percent)
[06:55:41] 3 (Run 35, Clone 34, Gen 66)
[06:55:41] 
[06:55:41] 34, Gen 66)
[06:55:41] 
[06:55:41] Entering M.D.
[06:55:48] Calling FAH init
[06:55:50] in POPC
[06:55:50] Writing local files
[06:55:50]  checkpoint)
[06:55:50] Read checkpoint
[06:55:50] 0 steps  (33 percent)
[06:55:50] PC
[06:55:50] Writing local files
[06:55:50] Completed 165000 out of 500000 steps  (33 percent)
[06:55:52] Extra SSE boost OK.
[06:55:54] - Couldn't send HTTP request to server
[06:55:54] + Could not connect to Work Server (results)
[06:55:54]     (171.64.122.86:8080)
[06:55:54]   Could not transmit unit 05 to Collection server; keeping in queue.
[06:55:54] + Sent 0 of 1 completed units to the server
[06:55:54] - Autosend completed

[Adak]

The few times that my router or modem has gone off the rails, I've had good luck with unplugging it, waiting for 45 seconds, and then plugging it back in.

It sounds like somebody shut down that port, all right. I've been turning in my WU's to 64.65.64:8080, without much problem, today.

Good luck!

[dabaerman]

Originally Posted by ThunderRd

@Ron: you must install under your windows admin credentials; that means that you MUST have a password to log into windows. If you don't have a password, go to user accounts and create one for the admin account. Then, uninstall everything you have done and start again, following the procedure I gave you again.

When the client configuration asks you for a credential store password, make it blank. But when MPI later asks you to enter your username/password twice, hit return for username, and then enter your logon password.

There is no way around this. The client WILL NOT RUN without a windows logon password.

TR, got that. I believe it was the uninstall part that I missed. will clean that one up later today. already have a 1900+ pointer/750 ppd on #1.


Adak, yes, taskbar! FAHmon works for monitoring the client, and will be happy to keep an eye on things for a while before running as a service.


thanks for all the help. not only will I get more points, the team and F@H will benefit.

Ron

[ThunderRd]

OK Ron, let me know if you need more help.

As for my folding problems, I am in the process of making sure that nothing has changed in my router configuration before going after my ISP with a 9mm.

I have to believe that somehow, something is wrong on their end. I have changed nothing intentionally. Haven't even installed any new software since the Chimp thingy - during which my boxen were steaming along quite nicely.

So the real fight prolly begins sometime tomorrow. I'll keep you posted. BTW, home machines not affected; but they are on a different ISP

[dabaerman]

TR, I now have SMP running on 4 of 5 boxen. 1 box is having issues. ie, create_credential_store
has encountered problems and needs to close.
I have deleted or removed (using add/remove) everything twice so far. no
luck. OS is Win2KPro. can do XP or Linux if necessary.

Ron

[ThunderRd]

Not sure about the win 2000; the download page says the client supports Windows XP/2003/Vista/2008. I'm not an expert, but would 2003 be the same?

I personally haven't run it on anything but xp and Linux. Linux is clearly the best choice for stability; I have 3 Linux boxes and they simply NEVER crash. Weeks at a time with no booting, folding 24/7.

The Windows client is much improved, but there are still stability issues from time to time. If you're not inclined to use/learn your way around Linux, install XP and go with it.

Are the 4 running boxes on XP? Or what OS? Don't fret, we'll get it going on the last one. 4 out of 5 ain't bad for a start Considering I did that how-to from memory

[dabaerman]

TR, 3 are XP 1 is 2KPro! the machine with 2KPro is a 939 dual core grinding on a SMP Gromacs 2100+ that will make the deadline.
if there are no server issues! I have a spare XP CD so will install that on #5.
when #6 is ready, Linux will go there. #5 almost done!

Ron

[ThunderRd]

Hmm. Can't say why #5 didn't like the client, but XP should. (Linux better though;: )

[ThunderRd]

I've edited and finished the installation how-to for the 5.92 client here:

Smp How to install the 5.92 beta client for SMP Folding

@Sam/Daniel:
I think it would be a good idea to sticky it so it stays up top; it's currently buried a couple of pages back in this thread and is a bit hard to find. Whaddya think?

[Samuknow]

I will put it in the folding section on the front page.

I really want to thank you for all your hard work. You have no idea how much it is appreciated.

[dabaerman]

SMP and Memory question. is SMP faster while running mem in Dual mode? I may be having memory issues, ie; one dead stick. only shows 1024 in bank 0 when 2048 are on the board. got more mem coming for the next box, that will be delayed now.

Ron

[ccperf721p]

Memory speed doesn't seem to effect the WU's much. The most I was ever able to get after beating up on my memory was an extra 25 ppd..

For the life of me I cannot get either 5.91 or 5.92beta to work under Windows. FILE_IO errors every time, I'm beginning to think the problem is at a hardware level. I've already had to sell 2 of my larger WD hard drives and buy hitachi drives due to the WD's not playing nice With my p5k.

I moved my linux install back to a 32 bit environment because I was tired of having to use double the space working on a project in a 32 bit chroot. I think I will have to stop foling until I finish up this project then do a clean install of both Linux and Windows then start over..

[dabaerman]

Originally Posted by ccperf721p

.. For the life of me I cannot get either 5.91 or 5.92beta to work under Windows....
I think I will have to stop foling until I finish up this project then do a clean install of both Linux and Windows then start over..

that bytes. it wasn't until I completely removed everything and anything to do with FAH, then reinstalled step by step, did it work for me.
I dislike reinstalls of OS. Winders most of all. as I get better with Linux, it reinstalls rather easy. good luck with the project and get back as fast as you can.

Ron

[ccperf721p]

I ran into the same problem on this rig awhile back, never did solve it and I spend most of my time at home in linux so I just let it be, figuring it was trace files or similar. This time around was on a fairly clean install with nothing f@h related on the machine, but everything else was installed again.

I actually believe it is the motherboard, after having a number of problems with WD drives in either RAID or AHCI mode, I rma'd the board only to have the same problems again. I borrowed a hitachi and the problem went away, except for the file_io errors in f@h. Last night I used the same drive with a repair install on a AMD single core rig and no errors.

I've had my eye on a dfi x38 board for awhile so maybe it is time to give dfi another chance at redemption.