It does not work on host ID 8882

Nightlord if it is not the wrong version of Boinc as just suggested, I'd suspend Boinc processing take another look at the client_state.xml file on that machine.

Between the:
<project>
<master_url>http://www.gpugrid.net/</master_url>
<project_name>GPUGRID</project_name>

and the

<duration_correction_factor>1.000000</duration_correction_factor>

we have been looking at you should find these two lines amongst the other stuff:

<nrpc_failures>0</nrpc_failures>
<master_fetch_failures>0</master_fetch_failures>

both need to be 0 - Burdett and I only needed to change the top one. It is just possible your second machine has a number greater than 0 for master fetch failures.

Phoneman1

Thanks Phoneman1, it's a Linux64 box, as is the other host that worked.

I just verified the client_state file, the nrpc_failures was at 3 and master_fetch at 0. I've played around with a few changes, but to no benefit. At present this host is dead in the water.

/edit: I detached and re-attached with no effect. I then backed up the Boinc installation and made a fresh one, connected to GPUGrid and it gives the same error. This still fails, now with a clean client_state file.

Time for a new Host ID.....

....... after initial success, I'm back to having problems getting WUs.
Ho Hum.

P.

OK, this is not funny any more :-(

With a new host ID 19709 and a clean installation of 6.4.2, it still gives me the same error messages.

I'm at a loss now.

My boxes are running since yesterday on 6.4.2 and I have set the DCF to 1. The two GTX260² and the GTX280 are running now without any problem, also with a estimated time of over 66 hours instead 7:28h. Only my 8800GT has WUs with a estimated time of over 252 hours, but I have 3 WUs. To get more WUs I use a trick: I manually update, then I get the 24 hours delay, and make a second manually update immediately. Then I get another error message (last call nn seconds) and the 24 hours delay is away. When now the rest time of the running GPU-WU is lower then the workcache (I have 1.5 days) I get new work automatically.

---

I have 5 machines running GPUGrid, 3x linux64 (fedora9) core2 quads and 2x winxp32 core2 duals, each has 1 GPU.
I don't think I've had a machine run out of WUs (yet!!).
All 3 linux machines usually manage to maintain at least 3 WUs whereas the 2 windows machines usually get to about 80% of their last WUs before downloading more WUs.
They all run 6.3.19 apart from 1 of the windows machines which runs 6.3.21 and they all run 3+1 (or 1+1 for the duals) and run Seti & RCN in addition to GPUGrid.

Their estimated completion times range wildly from a normal 2 hours upto +70 hours but this doesn't seem to affect their ability to get new work.

I don't know why they get work when a lot of you don't get any. I haven't edited any config files since installing 6.3.19 which removed my requirement to set ncpu-1. On the 2 machines here, the DCFs are 0.062 and 0.137 and have never been much higher that that.

I know a lot of you are running 6.4.x versions of boinc, so is it worth downgrading to 6.3.19 (or 6.3.21) to see if that helps?

I just finished my first WU. As soon as it reported, it got another.

Running WXP and 6.45 on a single core and a slow 8600 GT.

Hmm - I am running 6.4.5 with nVidia 180.60 and have two WU's on a 9800gtx+. Sorry about your problems. I did get that cell processor out of work for your system message this morning, but a reboot cleared it.

6.4.2 per today's recommendation. Downloaded 2 wu's by resetting the project and suspending everything else this morning. After finishing those 2 tonight, tried the same thing without success.

GTX 260, idle again. Server-side scheduler appears to be fubar.

12/18/2008 9:32:11 PM|GPUGRID|Sending scheduler request: Requested by user. Requesting 1468800 seconds of work, reporting 0 completed tasks
12/18/2008 9:32:16 PM|GPUGRID|Scheduler request completed: got 0 new tasks
12/18/2008 9:32:16 PM|GPUGRID|Message from server: No work sent
12/18/2008 9:32:16 PM|GPUGRID|Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
12/18/2008 9:32:16 PM|GPUGRID|Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.

http://www.gpugrid.net/show_host_detail.php?hostid=17947

6.3.x will at least allow me to DL work by resetting/suspending everything. 6.4.x seems to make it impossible to maintain work.

Nightlord, please send your client_state.xml at ps3grid@gmail.com.


Thanks.

gdf

From the server logs you are requesting 1 sec of work.

If version 6.4.2 is the preferred version, where can we get it. THe BOINC site only has the 6.2.19 and 6.4.5 available for download (as far as I have been able to push the system).

If version 6.4.2 is the preferred version, where can we get it. THe BOINC site only has the 6.2.19 and 6.4.5 available for download (as far as I have been able to push the system).

Here you can get all versions:

all versions of the BOINC-Manager

---

If version 6.4.2 is the preferred version, where can we get it. THe BOINC site only has the 6.2.19 and 6.4.5 available for download (as far as I have been able to push the system).

Hi Paul,

you can get it from http://www.gpugrid.net/join.php?sys=gpu or http://boinc.berkeley.edu/dl/.

---

pixelicious.at - my little photoblog

If version 6.4.2 is the preferred version, where can we get it. THe BOINC site only has the 6.2.19 and 6.4.5 available for download (as far as I have been able to push the system).

Hi Paul,

you can get it from http://www.gpugrid.net/join.php?sys=gpu or http://boinc.berkeley.edu/dl/.

Thanks ...

6.5.0 did not work, so let us see about 6.4.2 ...

Nope, no work for your type computer.

So, there is something else going on with the server regardless of the version of BOINC I run ...

{edit}I did run off 11 tasks for SaH last night and at least 3 of them have validated against a wingman... so, slight progress ... but GPU Grid is more like the kind of work I would want to do rather than just rip off SaH nonsense.{/edit}

6.5.0 did not work, so let us see about 6.4.2 ...

Nope, no work for your type computer.

When I saw an earlier post about 6.5.0 working well I decided to upgrade both my machines. As I wanted to try and determine what caused all the problems yesterday when it became necessary to reset nrpc_failures in the client_state.xml I thought I'd upgrade using differnet methods on my two machines. On the machine with the faster graphics card I upgraded BOINC the way I used to do it:

Open BOINC Manager
select task tab and check nothing is about to up load
select message tab and make sure no schedule requests outstanding etc.
Select activity / suspend
Select File / exit
Open task manager and check that all boinc; boincmgr and boinctray processes are not in the list.
Run the Boinc Installer

On the machine with the slower graphics card I cut the procedure down to:
Run the Boinc Installer

It looked as if this upgrade had worked but could I get a download of GPU on this machine with the slower graphics card - no.

I had to edit the client_state.xml file on this machine and change the nrpc_failures value back to 0. In fact I had to do this twice because I had not noticed Boinc had a client_state_next.xml file to replace the file I had just edited as soon as I restarted Boinc.

Once I had Boinc stopped and there was no "next" file in the directory (Boinc only creates one in certain instances and deletes it when it is not needed itself). I was able to edit the client_state.xml and restart Boinc.

At that point in the folder view I right-clicked on client_state.xml and selected copy. Then I right-clicked outside the file list and selected paste. Looking at the copy I had taken I could see that nrpc_failures was still 0 and sure enough going into Boinc and selecting GPU in the projects tab and pressing the update button brought me another wu.

It is just possible this "next" file business could be the cause of your problem today Paul and also Nightlord's problem from yesterday.

So the long winded way of upgrading worked for me today but the short-cut turned out to be anything but.

Yes I have 1 plus 4 tasks running in normal mode on both machines using 6.5.0 with estimates that look a lot better than anything else I've tried in the last three weeks with this project.

Phoneman1

Between the:
<project>
<master_url>http://www.gpugrid.net/</master_url>
<project_name>GPUGRID</project_name>

and the

<duration_correction_factor>1.000000</duration_correction_factor>

we have been looking at you should find these two lines amongst the other stuff:

<nrpc_failures>0</nrpc_failures>
<master_fetch_failures>0</master_fetch_failures>

both need to be 0 - Burdett and I only needed to change the top one. It is just possible your second machine has a number greater than 0 for master fetch failures.

Phoneman1

@GDF @Guys:

You have to know that - it really works. Phoneman u r great :). Both have to be 0!

I just installed new OS. After many big troubles with getting new WU on my maschines (with really different soft- and hard- ware configuration) i wanted to try new possibilities.

The new possibility was:

XP32 SP2
178.24 (old but checked and stable for sure)
6.5.0 - a fresh breeze ;)

But it doesn`t work. I mean again `no molecular` for me. Sure it was on fresh OS, fresh BM (it means DCF 1 etc.). I tried few times during some hours. Nothing. Then I`ve changed nrpc to 0. Manual update and voila :). But that`s not all. It was fantastic request (as always should be, but wasn`t never/ever in last weeks) which gave me 2 WU`s. After a moment BM requested again - next 2 WU`s arrived :). 1h23min - 1h58min remaining/WU.

Then I made experiment. After some of BM requests (after some minutes) I`ve changed nrpc again to 0 and with 4 GPUG WU`s (max.) and with a lot of Prime WU`s (not stopped! - Prime Resource share - 50%, GPUGrid Resource share - 100%) I made Manual update. Yes, there was correct request :).

Then next experiment. Little risky ;) :P. I`ve aborted one WU (some of you could get killed for one WU and I`m aborting it...!!! ;D ). And I made Manual update. Request correct, but got no new (the 4th one) WU, cause (...)molecular no available bla bla bla(...). Tried twice - same effect. Then I`ve changed nrpc again to 0 and manual update. Request correct, but... (...) molecular not available etc.(...) :O. Restarted comp, checked client_state, changed Resource share to 1000, changed DCF to 0.000001... Request 0... Request 0... Whatever I did, always request 0. Changed DCF back to 1.000000. Request correct, but (...)molecular not available(...). I can reach correct request, but there`s now way for me to receive any more WU`s.

Mby there`s something on server side, some note in database, that this maschine has received an error with one of WU`s and it means that this maschine should not receive new jobs. Or something similar.

Anyway I hope it will help some guys to get some WU`s and mby help you GDF to wonder WTF is going on now :). Remember, there was everything PERFECT, until first error. And you know that because of memory leaks probably every maschine will be at the same place as me now - mby go this way...

---

Well, I just tried setting nrpc failures to 0 (it was 3) but, not available for your computer.

I mean, I guess, I am fine if the project does not want my help ...

But, it sure seems that this is a server side problem (at least to me) and there is a curious reluctance to look there ...

I have done my part and have done what they have been asking.

GPU processing is intriguing, but, I guess I will have to wait for EaH to get it right. Or SaH to get Astropulse converted.

Trying and trying again. Every values correct now (DCF, nrpc etc.). Stopped other projects. Manual update. Got new WU`s :). - Guys, use it!

---

Sorry I was at work all day :-(

Client_state.xml just sent.

When I try to run 6.4.2 From this site I get
'Error in setup initialization file " on two systems... dont Know why. Is 6.4.2 the still the preferred version?... 6.4.5 is installed on my sys. now.

---

When I try to run 6.4.2 From this site I get
'Error in setup initialization file " on two systems... dont Know why. Is 6.4.2 the still the preferred version?... 6.4.5 is installed on my sys. now.

6.4.5 is the version with the biggest problems here on GPUGrid. Try the new developer version 6.5.0 or reload the 6.4.2. Be sure to clear the cache browser first, so that you can be sure to get a new download and not the broken file from cache.

---