climateprediction.net home page
Download Checksum Error

Download Checksum Error

Message boards : Number crunching : Download Checksum Error
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · Next

AuthorMessage
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 45764 - Posted: 29 Mar 2013, 15:02:16 UTC

Please DON'T post email addresses on these web pages.
Well, unless you really WANT spammers and their automated search programs to find them. :(

ID: 45764 · Report as offensive     Reply Quote
Profile mo.v
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 2363
Credit: 14,611,758
RAC: 0
Message 45765 - Posted: 29 Mar 2013, 15:29:54 UTC
Last modified: 29 Mar 2013, 15:32:33 UTC

Yes, what a nuisance about having the daily task quota reduced when it isn't your fault. Just for a moment's light relief I'll show you somebody who has no difficulty grabbing models.

Art, I wouldn't mess with the computer that has two models running on 7.0.55. Richard's working on the problem and if he needs extra evidence of problems he'll post back and ask one or both of you to do whatever is required. He's probably doing some testing of his own.

Take a backup of the complete contents of the BOINC Data folder now because if the problem turns out to affect uploads as well you may need to restore it. Exit completely from BOINC before copying for the backup.
Cpdn news
ID: 45765 · Report as offensive     Reply Quote
Art Masson
Avatar

Send message
Joined: 16 Oct 11
Posts: 254
Credit: 15,954,577
RAC: 0
Message 45766 - Posted: 29 Mar 2013, 15:32:23 UTC

Sorry...I've deleted my email addresses from my post.

I still don't "get" how to post an image (jpg) file which is local to my computer.
I'm obviously suffering from a dumb user error...

I understand the format is

but how do I refer to a local file?
I've tried

but that doesn't seem to work

Any help appreciated.
Art Masson
ID: 45766 · Report as offensive     Reply Quote
Art Masson
Avatar

Send message
Joined: 16 Oct 11
Posts: 254
Credit: 15,954,577
RAC: 0
Message 45767 - Posted: 29 Mar 2013, 15:36:12 UTC

Trying again

I understand the format is
(img]http://example.com/pic.jpg[/img)
but how do I refer to a local file?
I've tried
(img]file:///C:/eventlog1.jpg[/img)
but that doesn't seem to work

Any help appreciated.
Art Masson
ID: 45767 · Report as offensive     Reply Quote
Profile mo.v
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 2363
Credit: 14,611,758
RAC: 0
Message 45768 - Posted: 29 Mar 2013, 15:38:09 UTC
Last modified: 29 Mar 2013, 15:38:37 UTC

You can show files and photos from your own computer on the independent phpBB forum (the one that's down) but I don't think you can here. AFAIK on this forum you have to upload files or pics to a website or a dropbox and link to that.
Cpdn news
ID: 45768 · Report as offensive     Reply Quote
Art Masson
Avatar

Send message
Joined: 16 Oct 11
Posts: 254
Credit: 15,954,577
RAC: 0
Message 45769 - Posted: 29 Mar 2013, 15:50:19 UTC

OK, Thanks.
I'll just let it to the experts to fix the problem and assume someone will ask if any more info is needed.
Thanks so much for your help

Art Masson
ID: 45769 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 1 Jan 07
Posts: 943
Credit: 34,373,662
RAC: 4,704
Message 45770 - Posted: 29 Mar 2013, 16:57:48 UTC - in response to Message 45769.  

I've sent a report to the boinc_alpha bug reporting list, with copies for the CPDN staff.

There appears to be a incompatibility between what the CPDN server is sending out and what these late (v7.0.55 to v7.0.59) clients expect to receive. I expect it could be fixed by changing either the server or the client, but it would be very difficult for individual users to fix it on their home machines. So, either go back to a client version which is known to have worked for CPDN in the past, or wait for news that one party or other has changed the things under their control. The problem is in the area of downloading new tasks - it should be safe to run existing tasks, and to upload/report them when done.
ID: 45770 · Report as offensive     Reply Quote
Profile mo.v
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 2363
Credit: 14,611,758
RAC: 0
Message 45771 - Posted: 29 Mar 2013, 18:01:45 UTC
Last modified: 29 Mar 2013, 18:05:53 UTC

Thank you very much for your time, Richard. We know where we stand and what we have to do.

The BOINC main page:

http://boinc.berkeley.edu/download.php

is still offering 7.0.28 which works with our models that contain some errors server-end. As far as I can see this problem affects all current CPDN models and probably all versions of BOINC after 7.0.28. If you have running models, keep the BOINC version you've got until they finish and set CPDN in the Projects tab to No New Tasks.

Depending on the time it takes Andy to fix things in Oxford, and bearing in mind that it's a holiday weekend and next week he also has to grapple with the iframe exploit problem, you will have to decide whether to go back to 7.0.28 now or later. Any version after 7.0.28 is an alpha test version so you'll have to change version at some point.

Back up the complete contents of your BOINC Data folder after exiting fully from BOINC before changing version. If you have running models and by chance they don't reappear after the version downgrade you'll be able to delete the entire contents of the new data folder and copy into it the contents that you saved.

Do not try to back up and restore individual models. I've seen the instructions, which were for BOINC 6 or even 5, and they would fill at least two screens. Let's say this doesn't work. I repeat - we back up and restore the entire contents of the BOINC Data folder.

If you have 7.0.28 or any version lower than that and want to run climate models, do not upgrade your BOINC to a version higher than 7.0.28 until this problem is fixed.

Many thanks to WB and Art for reporting your problems and explaining things clearly.

If anyone else is suffering from the same failed downloads with MD5 checksum errors, especially if you don't know what to do, please post here.
Cpdn news
ID: 45771 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 1 Jan 07
Posts: 943
Credit: 34,373,662
RAC: 4,704
Message 45772 - Posted: 29 Mar 2013, 18:22:00 UTC

One small note for volunteers who crunch for more than one project. This affects people who use graphics cards (GPUs) for crunching, and who use 'Anonymous Platform' - usually optimised applications. SETI is the project which comes to mind, but there are others too.

Shortly after v7.0.28, there was a change made to the estimate that BOINC makes about the speed of your GPU. If you download (GPU, anonymous platform) work with v7.0.28 or before, and then upgrade to a later version of BOINC, the tasks may fail with 'maximum time exceeded'. New work downloaded with a new version of BOINC is fine.

I'd just suggest that you avoid crossing the boundary between v7.0.28 and later BOINCs more often than you strictly have to, especially if you use GPUs.

I haven't found which version of BOINC introduced this MD5 problem: there's nothing which obviously catches my eye in the list of changes (and the likely-looking one happened a long time before v7.0.28). But there might be one in the middle which would suit everyone until we get this fully sorted out.
ID: 45772 · Report as offensive     Reply Quote
Art Masson
Avatar

Send message
Joined: 16 Oct 11
Posts: 254
Credit: 15,954,577
RAC: 0
Message 45773 - Posted: 29 Mar 2013, 18:44:28 UTC

You guys are great. Thanks so much...time to take some time off for Easter now that we have some good work-around solutions (notably back migrating to 7.0.28) once any processing of new work-units is completed...and marking our clients to receive no new tasks).

Wondering what happens to work units that fail this way...hopefully they are not "lost" and are resent to other clients until there is a successful downloading!

One final note. I do have a machine (which is remote from my home and in Atlanta) which has been running 7.0.44. It is in fact processing 4 work units at this time, but I note in the log multiple download failures -- which I can only assume is the check-sum problem. I therefore suspect the problem was introduced with client version 7.0.44 or earlier.

Art Masson
St. Charles, IL
VP IT Strategy
Kellogg Company

ID: 45773 · Report as offensive     Reply Quote
WB8ILI

Send message
Joined: 1 Sep 04
Posts: 161
Credit: 81,421,805
RAC: 1,225
Message 45774 - Posted: 29 Mar 2013, 19:10:48 UTC

This is small data sample, but after downgrading to 7.0.44 I have downloaded two tasks successfully. I need at least 7.0.42 because of the app_config.xml capabilities introduced in that version for GPU's.

Both 7.0.54 and 7.0.55 show the following fix: "Selective upgrades of the core client based on md5 hash compares."

Maybe those "fixes" broke something else.
ID: 45774 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 1 Jan 07
Posts: 943
Credit: 34,373,662
RAC: 4,704
Message 45775 - Posted: 29 Mar 2013, 19:45:31 UTC - in response to Message 45774.  

Both 7.0.54 and 7.0.55 show the following fix: "Selective upgrades of the core client based on md5 hash compares."

May I ask where you're reading that from? I don't recognise the language from my usual BOINC reading, and in particular it doesn't appear in BOINC 7 Change Log and news. [edit - except for the special Android version of the client, which doesn't apply here. CPDN on a smartphone, anyone?]

My attention is lighting on

7.0.54 --> 7.0.55:
client (and maybe others): fix XML parsing bug when skipping large unrecognized elements.

7.0.52 --> 7.0.54:
MGR (and possibly other components): XML parsing fix.

but the BOINC code-searching tool seems to be broken at the moment, to add to our other woes.
ID: 45775 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 45776 - Posted: 29 Mar 2013, 20:42:14 UTC
Last modified: 29 Mar 2013, 20:44:48 UTC

Thanks for researching this Richard.
Subtle bugs like this are the bane of programmers.
ID: 45776 · Report as offensive     Reply Quote
WB8ILI

Send message
Joined: 1 Sep 04
Posts: 161
Credit: 81,421,805
RAC: 1,225
Message 45777 - Posted: 29 Mar 2013, 21:50:18 UTC

Richard Haselgrove -

The "fixes" I referenced are at -

http://boinc.berkeley.edu/dev/forum_thread.php?id=8203

Look at the 7.0.55 and 7.0.54 lists of fixes.

ID: 45777 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 1 Jan 07
Posts: 943
Credit: 34,373,662
RAC: 4,704
Message 45778 - Posted: 29 Mar 2013, 22:50:28 UTC - in response to Message 45777.  

Richard Haselgrove -

The "fixes" I referenced are at -

http://boinc.berkeley.edu/dev/forum_thread.php?id=8203

Look at the 7.0.55 and 7.0.54 lists of fixes.

That's the

BOINC 7 for Android Change Log and news

thread. The MD5s they're talking about there are used to find differences in the source code, and merge the work of two different teams of developers together. They're nothing to do with checking the integrity of climate data being transferred over the internet to our computers for further processing. (Same principle, but very different implementation).

On a different topic entirely, two HADcm3n models seen to have downloaded successfully to a v6.12.34 host while I was out at dinner. That seems to confirm that BOINC's 'new' XML parser is - may I say? - buggered, but the old one is still working. Other developers have pitched in to support that interpretation.
ID: 45778 · Report as offensive     Reply Quote
WB8ILI

Send message
Joined: 1 Sep 04
Posts: 161
Credit: 81,421,805
RAC: 1,225
Message 45779 - Posted: 30 Mar 2013, 0:22:10 UTC

Richard -

I guess I didn't look at the title close enough. I did a search on "md5" and found that thread. Sorry for the incorrect info.
ID: 45779 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 1 Jan 07
Posts: 943
Credit: 34,373,662
RAC: 4,704
Message 45789 - Posted: 2 Apr 2013, 17:32:44 UTC

Berkeley have made BOINC v7.0.60 available for download, specifically in the expectation of solving the CPDN / MD5 download problem. I've got it installed on a couple of machines for general testing - seems safe enough to install - but there are no CPDN models available to download at the moment, so I can't confirm that the CPDN problem is fixed yet.

Further news as we get it.
ID: 45789 · Report as offensive     Reply Quote
Art Masson
Avatar

Send message
Joined: 16 Oct 11
Posts: 254
Credit: 15,954,577
RAC: 0
Message 45790 - Posted: 2 Apr 2013, 17:38:37 UTC

Richard -- would it help for WB and I to load 7.0.60 on one of our machines now to help validate the problem is fixed (or not) when CPDN models become available?

Art Masson
ID: 45790 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 1 Jan 07
Posts: 943
Credit: 34,373,662
RAC: 4,704
Message 45791 - Posted: 2 Apr 2013, 17:54:13 UTC - in response to Message 45790.  

Richard -- would it help for WB and I to load 7.0.60 on one of our machines now to help validate the problem is fixed (or not) when CPDN models become available?

Art Masson

If you're prepared to live on the cutting edge, yes, it would probably help if more than one pair of eyes was keeping a watching brief.

Since I wrote that, I've independently heard from Andy Bowery that he has changed the CPDN server settings so that the problem should not recur when new work is generated. However, I suspect that if an old job fails (for any reason) on a volunteer's machine, and a replacement is sent out, that replacement might still have the trigger problem.

So, it would be useful not only to run v7.0.60, but to look carefully at any new task that arrives - assuming it downloads successfully, try to work out whether it's Andy's fix (newly created work) or David's fix (handle the old work properly) which has been effective in that particular case.
ID: 45791 · Report as offensive     Reply Quote
Art Masson
Avatar

Send message
Joined: 16 Oct 11
Posts: 254
Credit: 15,954,577
RAC: 0
Message 45792 - Posted: 2 Apr 2013, 18:03:41 UTC - in response to Message 45791.  

Thanks. I'll load 7.0.60 on one of my machines this evening.
Living Large here!

Art Masson
ID: 45792 · Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · Next

Message boards : Number crunching : Download Checksum Error

©2024 climateprediction.net