PDA

View Full Version : HELP with BOINC/YOYO



Bryan
08-31-16, 09:02 PM
Here's the problem: sometime last night (early morning) the Yoyo project went down. When I got up this morning I couldn't get to Yoyo with a browser. I had 9 crunchers running Yoyo prior to that time and turning in work ... daily stats proves that. Of course while the project was down I couldn't upload finished work or get new work.

All nine Windows crunchers will NOT contact the project either to update or upload. I have 3000+ WU ready to turn in. Any of the machines can contact Yoyo's website using a browser. I moved a machine to Linux and I was able to get work and turn it in.

I have stopped/started BOINC, rebooted the computer, reset the modem/router all to no avail.

The BOINC client will launch the browser and take me to the main Yoyo page, my account, or the forums. It just won't contact the project with an update or upload.

The upload URL is correct in the client_state file.

So quite frankly I'm at a loss ... 9 machines doing the same thing. Yes, network communications is enabled on all machines =)) They will contact other projects when update is clicked.

Does anyone have any ideas. I really really don't want to have to detach and reattach to the project.

Mumps
08-31-16, 09:37 PM
Anything insightful if you add the following to your cc_config on one of those systems?

<log_flags>
<http_debug>1</http_debug>
</log_flags>

Mumps
08-31-16, 09:45 PM
Another thought. Go to the Transfers tab. Could it be the file transfers are in Backoff and need a tickle instead of a simple project Update? If there are files listed there that show the Upload Retry is in Backoff mode, selecting them and hitting "Retry Now" may shake something loose.

Bryan
08-31-16, 10:27 PM
Another thought. Go to the Transfers tab. Could it be the file transfers are in Backoff and need a tickle instead of a simple project Update? If there are files listed there that show the Upload Retry is in Backoff mode, selecting them and hitting "Retry Now" may shake something loose.

I've done that one umpteen times without success. I've never had that fail but now it tries for X number of seconds and then goes back to the backoff mode.

I'll turn on the debug and see if anything shows up.

Bryan
08-31-16, 10:40 PM
It mainly shows it can't connect to the project but doesn't say why. It is using the correct URL for upload.



[fxd] starting upload, upload_offset 0
Started upload of ecm_xy_1472611370_C192_139_81_970_0_0
[file_xfer] URL: http://www.rechenkraft.net/yoyo_cgi/file_upload_handler
[network_status] status: online
[network_status] got HTTP error - checking ref site
[network_status] status: reference site lookup pending
[file_xfer] http op done; retval -107 (connect() failed)
[file_xfer] file transfer status -107 (connect() failed)

Then it comes up and says project maybe temporarily down but internet refernce site is okay.

So I have no idea what the http error is.

Mumps
08-31-16, 10:52 PM
Peek at this thread..

https://boinc.berkeley.edu/dev/forum_thread.php?id=10869

I'd say, maybe try some of the below options:

<http_1_0>1</http_1_0>

<max_file_xfers>2</max_file_xfers>
<max_file_xfers_per_project>1</max_file_xfers_per_project>

<http_transfer_timeout>20</http_transfer_timeout>
<http_transfer_timeout_bps>10</http_transfer_timeout_bps>

<file_xfer_debug>1</file_xfer_debug>
<http_debug>1</http_debug>
<http_xfer_debug>1</http_xfer_debug>

zombie67
09-01-16, 08:55 AM
Yeah, many BOINC servers are easily overwhelmed when trying to upload/report too many tasks at a time.

I think my default cc_config.xml has an entry in there to limit it to 20.

Mumps
09-01-16, 09:39 AM
Hopefully Bryan will post again with some of the stuff we tried last night while on TeamSpeak. Weirdest thing is that the Windows host can't communicate with YoYo, but other projects are fine. And a Linux Guest running under VirtualBox on that same host has no problems. While even a Windows guest on the host fails the same way. Hopefully Bryan will get some logs to post that may help in understanding the network traffic that's going on.

We've gotten past the max file transfers thought because he can't even contact the project servers to add YoYo as a project on a Windows host on his network. Even though a Linux Guest on the same Laptop can do so successfully. Just extremely bizarre.

Hopefully he remembers my suggestion to see if he can add it on his laptop if he takes it to Starbucks. :) (Sometimes all it takes is coffee!) :))

Bryan
09-01-16, 10:54 AM
Basically it gets down to 9 machines were running Yoyo ECM without problems when their server crashed. Since then no Windows machine can upload work. A clean BOINC install on one of the machines could not attach to Yoyo (says project unavailable). Bring up a Win7 VM on that machine and it will not attach. Bring up a Linux VM - no problem attaching, crunching, uploading. Bring up a different Linux VM and no problem. On other machines switching them to native Linux -- no problems. So it is a case that no Windows machine works but anything Linux has no problem. BTW, on any Windows machine the browser can get to the Yoyo website AND to the upload server URL.

The modem/router has been reset multiple times. All connections (wired and WiFi) except the 1 machine were disconnected.

@Mumps, I'll try to get to Starbucks this afternoon :D

Also, I have no idea which Windows Event log to look at. There are a gazillion shown in the Event viewer and none of them are flagged as being a problem.

Bryan
09-01-16, 04:49 PM
@Mumps, I don't know if the laptop would work at Starbucks but it did work at McDonald's :) It had no problem and immediately communicated with the project. So I came home and tied it into the network via WiFi and dang if it didn't contact Yoyo and would have attached. I still had 1 machine running Windows so I tried to upload the completed WU and BINGO, uploaded and then reported.

I switched each machine (1 at a time) back to Windows and uploaded everything. I got about 170k out of it.

I have absolutely no idea what the problem was/is. I had tried to upload earlier this morning and it wouldn't work. I had tried to attach the laptop and it wouldn't work. I think McDonald's WiFi put some magic into the laptop and then I brought it home. It's gotta be that!!

Anyway, I'm up and running but crunching Cosmo ... I can be in for the kill on LAF and CNT on that one instead =))

Thanks everyone for lending a hand with this very perplexing problem!!!!