Advanced search

Message boards : Graphics cards (GPUs) : check sum and signature errors

Author Message
Ross*
Send message
Joined: 6 May 09
Posts: 34
Credit: 443,507,669
RAC: 0
Level
Gln
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20626 - Posted: 7 Mar 2011 | 1:02:40 UTC

07/03/2011 1:51:23 p.m. GPUGRID [error] MD5 check failed for F270-TONI_KKAL2-12-F270-TONI_KKAL2-11-100-RND9694_3
07/03/2011 1:51:23 p.m. GPUGRID [error] expected 78ae4b727ac681841c20a8edc19bf5f5, got 872e47a2b17f9baf4680fd91cb89aba5
07/03/2011 1:51:23 p.m. GPUGRID [error] Checksum or signature error for F270-TONI_KKAL2-12-F270-TONI_KKAL2-11-100-RND9694_3
Hi
have tried serveral times to get new WUs but only get these that fail.
No long run WUs available
Cheers
Ross*
____________

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20627 - Posted: 7 Mar 2011 | 7:05:05 UTC - in response to Message 20626.

You have 2 tasks in progress now.

Did you upgrade from Vista to Win7?

Ross*
Send message
Joined: 6 May 09
Posts: 34
Credit: 443,507,669
RAC: 0
Level
Gln
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20628 - Posted: 7 Mar 2011 | 7:32:11 UTC - in response to Message 20627.

You have 2 tasks in progress now.

Did you upgrade from Vista to Win7?

Hi
same box running Win7 home 64 i7980x chip with 1 GTX 580 running at stock that I have been using for the last 6 months. Was running Nvidia 266.37beta but I downloaded 266.58 today. Seems to have fixed the problem ????
I have had 12 failed downloads and one error a couple of mintes after start.
I will be montioring the next 24 hrs to see if I get any more.
If I get more I will consder going back to older drivers.
Cheers
Ross*
____________

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20629 - Posted: 7 Mar 2011 | 9:26:00 UTC - in response to Message 20628.
Last modified: 7 Mar 2011 | 9:33:03 UTC

From the error messages there were file transfer problems. For example, when trying to transfer the F747-TONI_KKAL2-11-F747-TONI_KKAL2-10-100-RND3505_3 file from the server to your system the MD5 check failed error was generated. This basically means the file failed the MD5 integrity check and the download was rejected.

I don't know what caused these errors but my guess is it's client side; the tasks that failed to download to your system managed to download to other systems and some have been returned. It's likely that a system restart fixed the problem; resetting TCP/IP settings.

Ross*
Send message
Joined: 6 May 09
Posts: 34
Credit: 443,507,669
RAC: 0
Level
Gln
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20723 - Posted: 19 Mar 2011 | 23:32:39 UTC - in response to Message 20629.

Hi
after about a week of no errors re downloading I have had 9 in the last 2 days.
All are
<core_client_version>6.10.58</core_client_version>
<![CDATA[
<message>
WU download error: couldn't get input files:
<file_xfer_error>
<file_name>p34-IBUCH_7_mutEGFR_110315-1-p34-IBUCH_7_mutEGFR_110315-0-10-RND5046_3</file_name>
<error_code>-119</error_code>
<error_message>MD5 check failed</error_message>
</file_xfer_error>

all are on my Vista 64 boxes so far.
I have rebooted both boxes this morning to see if that makes a difference, but there are all on the same network and there all running on NV 266.58 GTX 580s
Is there any other settings that I need to look at?
Cheers
Ross*
____________

Dagorath
Send message
Joined: 16 Mar 11
Posts: 509
Credit: 179,005,236
RAC: 0
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20724 - Posted: 20 Mar 2011 | 0:09:25 UTC - in response to Message 20723.

The BOINC FAQ offers advice for error -119 in this section.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20726 - Posted: 20 Mar 2011 | 7:33:29 UTC - in response to Message 20724.

If the systems are all on the same network it might be the case that you need to restart the router. Turn it off, leave it off for 90sec or more, then turn it on and restart the systems. If you have other devices such as a switch, turn them off and on as well.

If that does not work test your general Internet activity is normal.

Ross*
Send message
Joined: 6 May 09
Posts: 34
Credit: 443,507,669
RAC: 0
Level
Gln
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20743 - Posted: 21 Mar 2011 | 8:37:31 UTC - in response to Message 20726.

Hi
Thanks for the info, Tried your suggestions, so far so good.
Reseting the project just lossed 8 hrs of crunching [my fault ]
Just a little concerned I have 10 Wus working/ready to run at this time but 14 in progress on the website.
Can I delete those that are now on my active list?
Cheers
Ross*

____________

Toni
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 9 Dec 08
Posts: 1006
Credit: 5,068,599
RAC: 0
Level
Ser
Scientific publications
watwatwatwat
Message 20744 - Posted: 21 Mar 2011 | 11:04:13 UTC - in response to Message 20743.

Perhaps also test your RAM (memtest86 or similar). Faulty chip used as a buffer may be corrupting data in-memory.

Richard Haselgrove
Send message
Joined: 11 Jul 09
Posts: 1626
Credit: 9,379,166,723
RAC: 18,990,592
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20745 - Posted: 21 Mar 2011 | 11:20:47 UTC - in response to Message 20744.

MD5 checksums are validated (or otherwise) immediately on download. We had a spate of faulty tasks recently at ClimatePrediction where (still unexplained) faulty MD5 values got computed on the server and stored in the BOINC database. When the tasks were sent out, the client computed the correct MD5 locally, and faulted the tasks - though there was nothing wrong with the actual files.

The BOINC message/event log tells the user what MD5 was expected (sent by the server) and was was calculated. Tools like winmd5 can be used to check MD5 values for downloaded files independently of BOINC, to work out which end the problem is happening at - and similarly for other OSs, of course.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20748 - Posted: 21 Mar 2011 | 17:52:50 UTC - in response to Message 20745.

Most data corruptions occur during downloads, rather than at Incryption or unzipping (CRC). No other user reports of download problems here would suggest it was a networking issue, this time.

I'm aware that some ISP's are content sniffing and targeting some download file types (including Zip files). Roughly 2weeks to 1week ago this was being tested in a big way in the UK, but it's a worldwide problem. The ISP's are basically targeting online gamers (P2P) and downloaders and imposing download restrictions because many ISP's are heavily oversubscribed.

This system belonging to Ross is showing 8 tasks in progress online, but would only have 4 in reality.

I expect the server remembers sending them, but the client rejected four and didn't tell the server.

Richard Haselgrove
Send message
Joined: 11 Jul 09
Posts: 1626
Credit: 9,379,166,723
RAC: 18,990,592
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20749 - Posted: 21 Mar 2011 | 18:38:11 UTC - in response to Message 20748.

Most data corruptions occur during downloads, rather than at Incryption or unzipping (CRC). No other user reports of download problems here would suggest it was a networking issue, this time.

I'm aware that some ISP's are content sniffing and targeting some download file types (including Zip files). Roughly 2weeks to 1week ago this was being tested in a big way in the UK, but it's a worldwide problem. The ISP's are basically targeting online gamers (P2P) and downloaders and imposing download restrictions because many ISP's are heavily oversubscribed.

This system belonging to Ross is showing 8 tasks in progress online, but would only have 4 in reality.

I expect the server remembers sending them, but the client rejected four and didn't tell the server.

I doubt the word "rejected" is appropriate in the last sentence.

The host is also showing a lot of "Error while downloading" - the one I spot-checked was

WU download error: couldn't get input files:
<error_code>-119</error_code>
<error_message>MD5 check failed</error_message>

which would match the thread title. That could be corruption at either end: but since the same files seem to have downloaded OK to other hosts when the task was re-sent, it does seem to be more likely client (or client's phone line) problems than server problems (in the CPDN case, resends failed as well).

The tasks which appear on the server list, but not in BOINC Manager, are a different problem. We tend to call them 'Ghost' or 'Phantom' tasks on other BOINC projects - they tend to happen when the sched_reply_[project].xml file is the one that gets corrupted in transmission: BOINC is notoriously *non* fault-tolerant in this respect (most other parts are well-protected by acknowledgement and retry handlers). But if sched_reply goes missing, there's nothing a project, or host, can do except wait for timeout or deadline and re-issue the task.

ISPs should empatically *never* be interfering with HTTP traffic on port 80 (which is what all BOINC projects, except WCG, use for uploads and downloads). I'd be interested to see chapter and verse, and better yet a source link, for your assertion.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20753 - Posted: 21 Mar 2011 | 23:54:47 UTC - in response to Message 20749.

Resetting the router did the trick WRT getting new tasks, without network errors.

Data was probably being sent over several different routes, some long, and packet losses kept occurring over the long hops/bad routes. Once the router was reset it picked up fresh DNS information with better routes. Many ISP's change DNS info quite often, but this require new IP info to be sent out, otherwise the client router will be using the same info it has until it expires (days).

The next problem was as you say the ghost tasks, a Boinc issue. In 5 days they should go away - timeout.

Off topic and on your assertion that ISPs should emphatically *never* be interfering with HTTP traffic - ISP's that use silent proxies port filter HTTP traffic (port 80), and many use deep packet sniffing in either so-called tests (40% of all users for years), or on all but the most expensive service. WCG uses 443 (HTTPS), but that doesn't mean ISP's cant restrict traffic in ways that interfere with this download/upload method. Some ISP's don't even know what traffic shaping/management they have in place at any given time. What is clear is that it results in widespread packet loss which nobbles uploads and downloads of all types. Bandwidth can be lowered, contention increased and routing over less favorable connections is also used. Sorry but you can't find such info in a single web page, and restrictions vary with different providers.

http://www.ispreview.co.uk/articles/10_Uncovering_ISP_Fair_Usage_Policies/01.php
http://www.pcpro.co.uk/news/broadband/365782/virgin-upstream-throttling-angers-gamers
http://virgin.net/allyours/faqs/trafficManagementFAQ.html
http://shop.virginmedia.com/help/traffic-management/traffic-management-trial.html
http://en.wikipedia.org/wiki/Virgin_Media
You can look up the other ISP's yourself.

Post to thread

Message boards : Graphics cards (GPUs) : check sum and signature errors

//