Worst bug ever: short time without internet => most/all tasks fail!!!

Message boards : BOINC client : Worst bug ever: short time without internet => most/all tasks fail!!!
Message board moderation

To post messages, you must log in.

AuthorMessage
dentaku

Send message
Joined: 14 Dec 06
Posts: 74
Germany
Message 14124 - Posted: 26 Nov 2007, 21:49:54 UTC

This is the worst bug of BOINC I ever encountered in a new version (Linux 5.10.28): when the internet connection is unavailable for a short time (e.g. 1 Minute), BOINC pops up a message that internet connection is down and you should choose the menu item extras > retry communication. But as soon as I press OK, most of my tasks suddenly change to 100% and with calculation errors. Grrrrr!

This occurs everytime when my internet connection is down and it's a pain in the ass to loose all this work (e.g. sevral days of crunching on Einstein, SETI, etc.). I can't see some rule here, as some tasks keep staying and continue and others fail. I have a dual core CPU running 2 WUs at a time and today, about 8 tasks failed on the last internet "pause". Too bad ... :-(
BOINC 7.2.42 (x86_64) on Linux Ubuntu 16.04 (64 Bit), AMD APU 7850K 3.7 GHz, 32 GB RAM.
ID: 14124 · Report as offensive
MikeMarsUK

Send message
Joined: 16 Apr 06
Posts: 386
United Kingdom
Message 14125 - Posted: 26 Nov 2007, 23:04:43 UTC
Last modified: 26 Nov 2007, 23:06:15 UTC

Sounds somewhat related to these trak tickets:

http://boinc.berkeley.edu/trac/ticket/113
http://boinc.berkeley.edu/trac/ticket/171 (duplicate of 113)
http://boinc.berkeley.edu/trac/ticket/282 (partially related)
http://boinc.berkeley.edu/trac/ticket/286 (partially related)


Could you scan through them and add your experiences to the most appropriate ticket? (With as much detail as possible). What was it that caused your internet connection to drop, was it internal or external to your PC? (i.e., a software firewall for example).
ID: 14125 · Report as offensive
Nicolas

Send message
Joined: 19 Jan 07
Posts: 1179
Argentina
Message 14126 - Posted: 27 Nov 2007, 2:05:46 UTC - in response to Message 14124.  

Probably this set of problems: message on the mailing list. (good that now I wrote a complete summary in a single place that I can link to, so I don't need to talk about it over and over again every time somebody asks)
ID: 14126 · Report as offensive
MikeMarsUK

Send message
Joined: 16 Apr 06
Posts: 386
United Kingdom
Message 14134 - Posted: 27 Nov 2007, 9:25:24 UTC
Last modified: 27 Nov 2007, 9:26:15 UTC


I do wish someone would revert the Boinc client back to async DNS, or simulate it in code. The problem that sync DNS was trying to solve seems to me to be a fairly minor issue (restart of Boinc fixes it), compared to the effect of having synchronous calls.

I know it would only be a partial fix, but it'd be a huge improvement.
ID: 14134 · Report as offensive
dentaku

Send message
Joined: 14 Dec 06
Posts: 74
Germany
Message 14190 - Posted: 3 Dec 2007, 19:45:12 UTC

As this seems to be a host of problems concerning internet connection and this single threaded/synchronous implementation also adds difficulty, I guess it won't be solved in short time. I wonder if there is a workaround ...
BOINC 7.2.42 (x86_64) on Linux Ubuntu 16.04 (64 Bit), AMD APU 7850K 3.7 GHz, 32 GB RAM.
ID: 14190 · Report as offensive
MikeMarsUK

Send message
Joined: 16 Apr 06
Posts: 386
United Kingdom
Message 14200 - Posted: 4 Dec 2007, 8:59:02 UTC
Last modified: 4 Dec 2007, 9:00:27 UTC

Some partial workarounds ...

* Revert back to 5.4.9 (uses asynch DNS. Some projects won't like this. Doesn't fix the other problems).

* Run 'network disabled' + a large cache for as long as possible (some projects with short deadlines won't like this. If localhost traffic fails you'll still lose your workunits).

* Change the local network architecture (router with integrated DNS, hardware firewall, costs money)
ID: 14200 · Report as offensive
Nicolas

Send message
Joined: 19 Jan 07
Posts: 1179
Argentina
Message 14207 - Posted: 4 Dec 2007, 17:25:20 UTC - in response to Message 14200.  

* Revert back to 5.4.9 (uses asynch DNS. Some projects won't like this. Doesn't fix the other problems).

An old 5.8.x will do; I don't remember what exact version changed DNS though.

* Run 'network disabled' + a large cache for as long as possible (some projects with short deadlines won't like this. If localhost traffic fails you'll still lose your workunits).

If your localhost traffic fails you have different problems...
ID: 14207 · Report as offensive

Message boards : BOINC client : Worst bug ever: short time without internet => most/all tasks fail!!!

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.