| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 26
|
|
| Author |
|
|
a_mobile_humanist
Cruncher Joined: May 20, 2011 Post Count: 34 Status: Offline Project Badges:
|
Well the driver idea is probably not applicable to me : i'm using a Desktop with a direct wired connection to my box... Like you Azriel, I'm using a wired connection... Probably should have checked that; everything else (archlinux, 64-bit, SSL) fit so nicely... ![]() |
||
|
|
a_mobile_humanist
Cruncher Joined: May 20, 2011 Post Count: 34 Status: Offline Project Badges:
|
...I can't see any packages that would have caused this problem at the date of the 7th of september... How are you checking the last update for the openssl and other packages? I ask because I'm downloading an Archlinux ISO now and note that the [Edit 1 times, last edit by a_mobile_humanist at Sep 10, 2012 10:06:18 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'm checking from the package listings on the Arch website: http://www.archlinux.org/packages/core/x86_64/openssl/, it says last updated 2012-05-11.
I went through every update in the 'recent updates' list from now and through to the 6th of September, but nothing looks like it has been updated that may effect BOINC connectivity, still I have a feeling it's OpenSSL problems. |
||
|
|
Azriel
Cruncher France Joined: Feb 10, 2008 Post Count: 5 Status: Offline Project Badges:
|
My log literally fails at around 21:07 BST on the 7th, when a few minutes before a download had successfully completed. Heres a short snip from my log: http://pastebin.com/K5ZqJiKz, the time it first fails is infact 21:07, just minutes before at 21:05 the last downloads I have received to date, something happened around that time I think. Woh that's an important message I missed here, so I went through my log and found this part which seems relevant: 07-Sep-2012 08:16:24 [World Community Grid] Started download of 75f9660976be7b87b22568d4e01ea0ee.dat.gzb A few seconds later, another download started ( X0960067810623200604261923_X0960067810623200604261923.jp2) and succeded, and as far as I can tell it was the last one to succeed. As I suppose the date are in local time, I should precise that I am living in Paris, and with Daylight Saving Time I'm currently at GMT+2. I've tried to look up BST on wikipedia to cross-reference with you, but there's apparently at least 5 points around the world using this acronym :p. But the bug seem to have happened to us within half a day apart. Funnily enough, according to /var/log/pacman.log i updated my system the very same day at 03:52 AM, so 3 and a half hour before the bug, and then nothing until 8:45AM, so 30 minutes after the bug... It's starting to look like it's not an update issue on Arch side. Is there a WCG tech here who could tell us if something was changed server side ? [Edit 2 times, last edit by Azriel at Sep 10, 2012 10:58:53 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'm in London UK, so my timezone would be GMT+1 (so we're pretty close), sorry about the confusion with BST.
It doe's seem more like server side change, I went through Arch updates again, the only packages that updated around that time, that are related to OpenSSL is MySQL, which shouldn't have effected us, and it isn't needed for BOINC. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Just an update, I experimented a little and 'temporarily' renamed my OpenSSL libraries (libssl.so.1.0.0 and libcrypto.so.1.0.0) to '.old', and created links named after them to OpenSSL 0.9.8 (libssl.so.0.9.8 and libcrypto.so.0.9.8) instead, I restarted BOINC, and at once work units downloaded freely. It works with 0.9.8.
----------------------------------------So I can confirm it is an OpenSSL issue and seems it is related to the 1.0.1 bug: https://bbs.archlinux.org/viewtopic.php?id=138103. The only problem I have is that BOINC downloaded WCG units before the 7th of September, I'm thinking something was updated during maintenance which is triggering the bug with OpenSSL 1.0.1c, even with the workaround flag. In any case, the current OpenSSL 1.0.1 patch and flag workaround doe's not work now in BOINC. Another discovery, the provided download from WCG 'BOINC client version 6.10.58 for linux(x86)', also works with no workarounds, it looks like it was compiled with OpenSSL 0.9.8g, but seems to work with the my Archlinux systems lib32-OpenSSL 1.0.1c. [Edit 1 times, last edit by Former Member at Sep 11, 2012 1:33:11 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Could be coincidence, but last week there was a cache corruption on the servers which started hitting machines on the 6th/7th. This cache was cleared/reset on September 10. See this thread. https://secure.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=391549
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
That would explain why these problems are occuring, although I still think OpenSSL 1.0.1 is the root cause. I followed the link and tried again using the information there, unfortunately, its still a no go, and using the older OpenSSL 0.9.8 libs got it to download, but I have to revert to the Archlinux default of 1.0.1c otherwise other packages will break.
I did have an application suffer with a bug that effected users of OpenSSL 1.0.0 and above, when session id's are enabled server side, the server drops the connection after the initial 'hello' message, the workaround was something like clearing the SSL session id cache using curl, but this would require a patch applied to BOINC directly most likely, I'll carry on looking and keep it updated here. |
||
|
|
RetiredTech
Advanced Cruncher Canada Joined: Feb 2, 2012 Post Count: 91 Status: Offline Project Badges:
|
I too cannot download work units. I'm running Windows 7 on an Intel machine.
----------------------------------------Event Log after BOINC Reset: 11/09/2012 4:25:11 PM | World Community Grid | Started download of ef0c409bf5be67870c7f7717592665af.dat.gzb 11/09/2012 4:25:13 PM | World Community Grid | Temporarily failed download of ef0c409bf5be67870c7f7717592665af.dat.gzb: transient HTTP error 11/09/2012 4:25:13 PM | World Community Grid | Backing off 7 min 31 sec on download of ef0c409bf5be67870c7f7717592665af.dat.gzb 11/09/2012 4:25:16 PM | | Project communication failed: attempting access to reference site 11/09/2012 4:25:18 PM | | Internet access OK - project servers may be temporarily down. Try again gets the same messages. OK! I aborted the stuck / failed download task and was rewarded with a deluge of work units! [Edit 1 times, last edit by RetiredTech at Sep 11, 2012 8:58:36 PM] |
||
|
|
Azriel
Cruncher France Joined: Feb 10, 2008 Post Count: 5 Status: Offline Project Badges:
|
@RetiredTech well, glad that you're able to download new WU ;)
----------------------------------------Little update on my part too, I've done the same trick as peaceseeker - renaming and symlinking - and I've also been able to downloads new tasks. I profited of the occasion to buffer 3 days worth of work, so at least my machine won't be idle while we keep fixing this thing ;) Also it'll be interesting to see if finished tasks are able to upload correctly - I seem to recall that they do, but it never hurts to double check. I also decided to post a thread on the Archlinux forums since it seems to be Archlinux-related, or at least GNU/Linux-related. Plus it might bring awareness to some users who just keep BOINC in the background and don't trouble about it so they might fix it too once a solution has been found. [EDIT] Well, reports seems to be uploading fine, so I assume that it's really a one-way problem... [Edit 2 times, last edit by Azriel at Sep 12, 2012 9:50:11 AM] |
||
|
|
|