Hi all, starting from a couple of days ago one of my computers (http://www.gpugrid.net/show_host_detail.php?hostid=178360) all the gpugrid tasks error out with this message

Looking at the 'workunit' column for those errors, every machine which has attempted them has failed right at startup. That looks like a bad batch of work, rather than any problem with your machine.

Whether it's related to the licence warnings, I can't say.

Edit - having said that, the machine in front of me has just started running gpcrmd_3nya_complex_withNA_rep_3-GIANNI_CPXB-8-10-RND9513. It's running fine, 2.5% in after half an hour. The first attempt at that workunit (under Linux) failed with the same error as you: I'm running Windows 7, and I'm careful to stick with a known good driver.

Edit again - the vast majority of the failures among your wingmates are also running some flavour of Linux. That feels like a Clue.

I've had 7 of these in a row since the 30th. Only 1 previous bad WU prior and nearly 100mil points worth of good tasks. System has been crunching away at E@H in the mean time.

System is a 1950x with GTX 1070 and 1070 Ti on Ubunutu 17.04.

All others who received my error'd tasks next are running Windows.

After a long absence from GPUGrid, I tried getting some work yesterday for my linux rig with 2-GTX1080 GPUs. After several hours of trying to get work, I did get one work unit and it failed with the same error code as posted above. Another computer with windows as OS finished that WU sucessfully.

http://www.gpugrid.net/workunit.php?wuid=13112686

I also notice on the bottom of the server status page that the failure rate for project work units vary from 13 to 20 %. Is that typical and are most of the failures on linux?

http://www.gpugrid.net/server_status.php

My computer:

3930K w/ 32 GB RAM
2-GTX1080
Linux Mint 18.3
Nvidia driver version 384.111

What is the recommended driver version for linux?

I've been folding successfully on this computer for a couple of months.

After a long absence from GPUGrid, I tried getting some work yesterday for my linux rig with 2-GTX1080 GPUs. After several hours of trying to get work, I did get one work unit and it failed with the same error code as posted above. Another computer with windows as OS finished that WU sucessfully.

http://www.gpugrid.net/workunit.php?wuid=13112686

I also notice on the bottom of the server status page that the failure rate for project work units vary from 13 to 20 %. Is that typical and are most of the failures on linux?

http://www.gpugrid.net/server_status.php

My computer:

3930K w/ 32 GB RAM
2-GTX1080
Linux Mint 18.3
Nvidia driver version 384.111

What is the recommended driver version for linux?

I've been folding successfully on this computer for a couple of months.

Here as well! The last 4 WUs haver errored out after a few seconds with:
process exited with code 212 (0xd4, -44)

Linux Lubuntu 17.10, AMD Ryzen 1700x and GTX1070. This system has worked very well before.

Edit again - the vast majority of the failures among your wingmates are also running some flavour of Linux. That feels like a Clue.

I've had 27 consecutive failures since Jan 30 on three different Linux machines (GTX-1060). Review of the results in the wingman column indicates all failures to be from v9.14 (ACEMD and Linux version?) and the successful completions using v9.18 Windows version I assume.

Edit: spelling

Has anyone sent a PM to Gianni to suggest he deprecates the Linux app (to save wasted bandwidth) until the licence problem has been sorted out?

Has anyone sent a PM to Gianni to suggest he deprecates the Linux app (to save wasted bandwidth) until the licence problem has been sorted out?

Up to 17 failed tasks for me. The project out put is really low. Looks like a test to see if the admins actually give 2 cents.

http://stats.free-dc.org/stats.php?page=proj&proj=ps3

Has anyone sent a PM to Gianni to suggest he deprecates the Linux app (to save wasted bandwidth) until the licence problem has been sorted out?

I haven't.
The better way to address this issue to put the question this way:
Who will send a PM to Gianni to suggest he deprecates the Linux app?
Perhaps they have already noticed this issue on their own Linux machines?

Hi Gianni,

We're discussing (in process exited with code 212 (0xd4, -44)) that all tasks run on Linux machines are failing. It started with your current tasks, but it may be more widespread than that.

One user (opening message 48771) has reported error messages relating to licensing expiry.

Windows applications are unaffected by this problem.

Thanks for sending the PM Richard.

This is the reply I got from the Acellera Help Desk

| Dear Keith,
|
| Thanks for your email and our apologies for the inconvenience.
|
| First, I would like to understand correctly what happened.
| May I ask you if you were running on GPU Grid?
| ACEMD basic license does not allow to run on GPUGrid, it is a different compilation.
| Then, We should be able to answer your request.
| Thank you for your understanding!
|
| I take advantage to ask you the following:
| Could you please let me know if you work for non profit entity?
|
| Finally, you can download free of charge the latest version of ACEMD at: https://software.acellera.com
| If you use more than one GPU, you may be interested by the PRO license.
| This license allows to run on e.g. cluster.
| Would it be useful for you ?
|
| Expecting your answer, I wish you a great day,
|
| Franck
`---

I believe Gianni is the scientific founder of Acellera so I would think he would get a steep discount on licenses.

I believe Gianni is the scientific founder of Acellera so I would think he would get a steep discount on licenses.

Somebody must be listening as I haven't had a gpu wu since my last post here on any of my 3 gpu capable machines. Just crunching QC, WCG and E@H right now. Also noticed I haven't received an upgrade to ACEMD v 9.14 yet (original download date 5/6/17 still on file).

Keep an eye on https://www.gpugrid.net/apps.php for new versions. That standard BOINC url works here as on all BOINC projects, even though there isn't an obvious link.

Keep an eye on https://www.gpugrid.net/apps.php for new versions. That standard BOINC url works here as on all BOINC projects, even though there isn't an obvious link.

Rats... I am getting more gpu work and they are all failing. Since I don't have any windows machines, I am out of luck until this is resolved.

Got same error code on my Linux box for several tasks:

http://gpugrid.net/results.php?hostid=447186

Most recent task, I suspended the task before it began. I rolled my computer's date back a month (February to January), then resumed the task to let it begin.

It didn't stop with an error immediately, that's a good sign.

I waited a minute, then returned my computer's date to real time. It's still running, hope it finishes okay.

http://gpugrid.net/workunit.php?wuid=13116635

More than half the current research projects on the server status page are now showing amber for the error rate. That's the sort of signal that might attract the admins attention.

I've had no response to my PM, but if there's no sign of change by tomorrow morning I might try again.

http://gpugrid.net/result.php?resultid=16993864

It was going so well, but suspending and resuming with the date set to February caused the error again. The app seems to check the date at the beginning, and upon resuming as well. Ten hours of work wiped out, d'oh.

Trying again, starting with a January date and then trying to never suspend. Maybe it'll work this time.

http://www.gpugrid.net/workunit.php?wuid=13117301

BOINC never keeps a task scheduled for a GPU in memory (there's no equivalent of paging or a swap file for graphics memory), so every 'resume' is actually a complete application relaunch. Try to avoid letting it swap out.

Success! So fiddling with the date is a possible workaround for Linux users.

http://www.gpugrid.net/workunit.php?wuid=13117301

Edit: P.S. Suspend other tasks before rolling the date back. It's okay if they stay in memory, but if they're running, the rollback will lock them up. There are probably other ways to do it, but to be safe: I suspended everything, then rolled back date, then started GPUGRID task (manager won't show it running, but it is running), then corrected the date (manager now shows GPUGRID running), then resumed other tasks, and made sure the GPUGRID task never suspended.

Got a response to my trouble ticket at acellera.com about the app error from Gianni.

the error message is misleading, there is no license to request. We are already fixing the application in gpugrid.

Thanks for the reply Gianni. Any estimate when that will be made available?

not yet, but it will be posted on gpugrid forum. We have a problem that the person that builds it is out until the end of the week. We have to see if the others can do it without him.

Looks like the same issue with the Short Acemd tasks. Thought it might use a different application, guess not. Same error in Linux as before.

Will have to wait for the developer to fix the application.

Should we turn off GPUs in the meantime to avoid failed WU or it doesn't matter ?

Should we turn off GPUs in the meantime to avoid failed WU or it doesn't matter ?

Just completed a WU for the new version 9.19, seems to be working fine.

I've already done and validated two new tasks with the 919 application. Original problem is resolved.