| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Client hung not uploading or downloading after network disconnect |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 6
|
|
| Author |
|
|
Punchy
Advanced Cruncher Texas Joined: Nov 30, 2010 Post Count: 60 Status: Offline Project Badges: 
|
I noticed that one of my clients had not uploaded any work in the last few hours so I went to check it and the network cable was disconnected. All of the work on it is 100% complete and waiting to be uploaded. After verifying that internet connectivity was established, I tried to force uploads/downloads using the advanced/do network communication option. Nothing happens. All options are set to allow computing and networking always. No uploads, no downloads.
----------------------------------------In the tasks view all items are listed as "uploading". In the transfer view they all show 0% transferred, upload pending, project backoff 45 minutes and counting down. The backoff time started at 1 hour and I've spent the last 15 minutes trying to get things started, including restarting the manager. It's very frustrating having this system idling for at least the next 45 minutes doing nothing. Is there any way to break this logjam and get the system productive again? If it weren't going to cause the loss of an entire days' work I'd already have uninstalled and reinstalled by now.  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi Punchy,
----------------------------------------Sometimes we make problems bigger then they are ;D ... In the transfer tab, did you select any file to upload and then hit the "Retry Now" button on left? That will force a new upload attempt. It's one of multiple steps described in a Start Here FAQ. When Client-Server Communications Started to be Troublesome http://worldcommunitygrid.org/forums/wcg/viewthread?thread=21569 99 out of 100 working through these makes things move again. The "Do networking" on the menu simple tests if the client is connected, but does not actually reset any backoff counters. --//-- [Edit 1 times, last edit by Former Member at Dec 17, 2011 3:29:41 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I just hit "Update" on the Projects tab.
|
||
|
|
Punchy
Advanced Cruncher Texas Joined: Nov 30, 2010 Post Count: 60 Status: Offline Project Badges:
|
Rob, thanks for the suggestion. You may already know this, but the issue I have with the FAQ here is that it is SO extensive it makes answers hard to find. If a keyword search won't find the FAQ I probably won't see it.
----------------------------------------I was unable to find a quick solution here with a search on the word "backoff" but I did find several references at Berkeley's site. It appears to be a common networking issue with the client. A quick reboot fixed the problem. "Update" didn't work either - I would just get a message "Not requesting or uploading tasks" or something to that effect. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Booting always looses some progress on work underway [and can loose hours on CEP2], so your first step was "Retry Now" as noted in my previous post.
I'll think about creating an additional Network Issues sub-section inside Chapter 4 "BOINC Messages and Problem Resolution" and move/copy anything network related topics in there with some additional keywords. If one word does not pop up, think of the root issue... communication, network. Usually there will than be a hit. Some topics are actually in the index under multiple descriptions to increase the chance when doing a "search webpage while typing". --//-- |
||
|
|
Punchy
Advanced Cruncher Texas Joined: Nov 30, 2010 Post Count: 60 Status: Offline Project Badges:
|
There was no progress to lose as all downloaded work was already at 100% completion, so that didn't seem like much of a risk.
----------------------------------------I think it would be helpful to add as many of the unique error message words as possible to increase the hits on the FAQ. I typically wouldn't use network or communication as a search term as they are too common, but backoff is pretty unique, or "not requesting or sending work" (or whatever the specific message was). I thought "Do network communication" actually did cause things to happen. Using "update" plus "do network communication" seems to be the only way for me to queue up extra work before an extended network downtime. On a 24-thread machine, setting the additional work buffer to a day or two, it still seems to only download 15 extra workunits unless I keep forcing update/do network over and over. |
||
|
|
|