Project files stuck in download

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2143
Credit: 3058316249
RAC: 2094571

ES-Six wrote: I tryed on a

ES-Six wrote:

I tryed on a linux computer and I encountered the same donwload errors but not on the same files (probably because as it's a linux computer, required files are not the same) :

The program files are different, but the data files are the same. And it's data files that are missing.

ES-Six wrote:

Note : on both machines I added the project so it seem's to affect every newly attached hosts :)

More likely, it's just that newly attached hosts are allocated newly generated data.

ES-Six
Joined: 6 Oct 18
Posts: 18
Credit: 30970261
RAC: 1653

Richard Haselgrove

Richard Haselgrove wrote:

ES-Six wrote:

... an einstein@home server that is returning an HTTP error 404 when the BOINC client is attempting to download required files (specifically NVidia GPU and tasks related files).

Have a nice day ...

As you will see for yourself if you click on the links in my message 192810. Doesn't matter whether it's BOINC or a browser, Windows or Linux - if the file ain't there, it ain't there.

 

Ok, I had some download errors for files with this kind of names on my windows machine, it confirm there is also tasks file download that is affected by the bug.

I tested to download files throught my web browser too and got this same 404 error so yes, files are not there, but they should be there.

I also tested to change the base domain to the following doamain : einsteinathome.org and it works for some files (binaries) but not for tasks files.

San-Fernando-Valley
San-Fernando-Valley
Joined: 16 Mar 16
Posts: 567
Credit: 10962709465
RAC: 11423943

ES-Six wrote: For privacy

ES-Six wrote:

For privacy reason, I hide all my computers from public users, so when you will finish to look at my specs I will hide them again.

Sorry, I thought you had not noticed that the PCs are hidden. I'm finished.

Strange errors.

I have no further ideas - I'm just glad all of mine work!

Cheers ....

ES-Six
Joined: 6 Oct 18
Posts: 18
Credit: 30970261
RAC: 1653

Richard Haselgrove

Richard Haselgrove wrote:

ES-Six wrote:

I tryed on a linux computer and I encountered the same donwload errors but not on the same files (probably because as it's a linux computer, required files are not the same) :

The program files are different, but the data files are the same. And it's data files that are missing.

ES-Six wrote:

Note : on both machines I added the project so it seem's to affect every newly attached hosts :)

More likely, it's just that newly attached hosts are allocated newly generated data.

Yes program files are differents as the binary format is operating system and architecture related (x64, arm, etc.).

 

I think to potentially 2 things that can explain this problem :

1) A hard drive on one of the einstein@home server is failing and some files are unreadable

2) A software update for boinc servers messed up something (e.g. bad generated URL in the project configuration) :

- I also tryed with BOINC 7.16.16 and BOINC 7.16.20 and the problem appear for both versions of BOINC so it's not related to the BOINC clients. And last year, I used the same computers(and some more) to contribute without any issues.

3) Workunit generation script is failing and outputing bad download path (but in this case it doesn't explain the 404 download error for binaries)

4) Also a possibility but I don't think it's this one as I'm not alone to encounter download errors and I'm also not in the same country : network HTTP requests could be intercepted because they are not HTTPS so they are vulnerable to MITM attacks.  (eliminated, see below)

5) Human error (unexpected file deletion on the server hard drive)

Edit : I just used some proxies to test why files are not available and : I can eliminate the point 4) because the tested file is unreachable in all cases :

tested file download link : http://einstein2.aei.uni-hannover.de/download/1c9/h1_0635.00_O3aC01Cl1In0

ES-Six
Joined: 6 Oct 18
Posts: 18
Credit: 30970261
RAC: 1653

San-Fernando-Valley

San-Fernando-Valley wrote:

ES-Six wrote:

For privacy reason, I hide all my computers from public users, so when you will finish to look at my specs I will hide them again.

Sorry, I thought you had not noticed that the PCs are hidden. I'm finished.

Strange errors.

I have no further ideas - I'm just glad all of mine work!

Cheers ....

No worries,

Computers are now hidden.

Cheers :)

ES-Six
Joined: 6 Oct 18
Posts: 18
Credit: 30970261
RAC: 1653

I investigated more :I

I investigated more :

I read more HTTP debug logs of my boinc clients and I observed links and domains used :

I notices the client attempt to download files from two domains einstein-dl.syr.edu and einstein2.aei.uni-hannover.de domain.

Exemple :

http://einstein-dl.syr.edu/download/174/h1_0636.40_O3aC01Cl1In0

and

http://einstein2.aei.uni-hannover.de/download/8c/h1_0635.80_O3aC01Cl1In0

 

So If I take a link not working because it returned 404 error and I change the domain like this :

http://einstein2.aei.uni-hannover.de/download/1c9/h1_0635.00_O3aC01Cl1In0

becomes

http://einstein-dl.syr.edu/download/1c9/h1_0635.00_O3aC01Cl1In0

I got a timeout error when trying to access file from a web browser.

When I ping einstein-dl.syr.edu I got a timeout too.

So : one of the einstein@home download server is down.

 

That left us with these options :

1) Einstein@home staff is making (unplanned) changes to the download servers (e.g. to replace the server) but it doesn't go as planned.

2) A software update for boinc servers messed up something (e.g. bad generated URL in the project configuration) :

- I also tryed with BOINC 7.16.16 and BOINC 7.16.20 and the problem appear for both versions of BOINC so it's not related to the BOINC clients. And last year, I used the same computers(and some more) to contribute without any issues.

3) Workunit generation script is failing and outputing bad download path (but in this case it doesn't explain the 404 download error for binaries)

4) Human error

 

Server testing results with a buggy file download link :

Server Respond to ping GET HTTP / succeed File download succeed (for a specific file)
einstein-dl.syr.edu [timeout] (may be NOT intentional, server is down) NO [timeout] NO [timeout]
einstein2.aei.uni-hannover.de [timeout] (may be intentional, server is up) YES [GOT redirection page to https://einsteinathome.org] NO [HTTP 404 error]

 

I think there is one main server and one mirror server.

In both cases, it's not a normal behaviour so there is definitely a problem on the einsteinathome.org server's side.

Is there an admin here?

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2143
Credit: 3058316249
RAC: 2094571

I don't think is as simple as

I don't think is as simple as 'single server down'. I have several machines which are still downloading and processing GW tasks as normal - currently working on frequencies between 0627.20 and 0637.20 (the next group of figures after the h1_ prefix). I think that the problem tends to pounce when we've finished all the tasks for a given frequency band, and the project tries to send us the data files for a new frequency band. If they're not ready, or have gone awol - boom!

With regard to the URLs: Einstein@Home has multiple servers around the world. In Europe, I get sent

<file><br />
    <name>h1_0635.80_O3aC01Cl1In0</name><br />
    <nbytes>17582296.000000</nbytes><br />
    <max_nbytes>0.000000</max_nbytes><br />
    <md5_cksum>4f85b13047f9f20465a174337ff51311</md5_cksum><br />
    <status>1</status><br />
    <sticky/><br />
    <download_url>http://einstein2.aei.uni-hannover.de/download/8c/h1_0635.80_O3aC01Cl1In0</download_url><br />
    <download_url>http://einstein-dl.syr.edu/download/8c/h1_0635.80_O3aC01Cl1In0</download_url><br />
</file>

with the European server (hannover.de) listed first. I think that people further west - in the Americas - get directed to (?) Syracuse University (syr.edu) in preference. The data files on both servers should be kept as exact mirrors of each other.

The project administrators don't really have time to sit and read these messages. They usually rely on the forum moderators to alert them to developing problems. I had hoped that Gary Roberts might have picked up on this one: if he hasn't commented by tomorrow, I'll reach out to Bernd myself.

mikey
mikey
Joined: 22 Jan 05
Posts: 12958
Credit: 1884513703
RAC: 18024

ES-Six wrote: I

ES-Six wrote:

I investigated more :

I read more HTTP debug logs of my boinc clients and I observed links and domains used :

I notices the client attempt to download files from two domains einstein-dl.syr.edu and einstein2.aei.uni-hannover.de domain. 

Do you use a VPN? If so try telling it you are in the US and see what happens, Try Canada, South America etc etc as since it works for some people you just need to make it connect to those Servers instead of the 2 that are failing for you.

ES-Six
Joined: 6 Oct 18
Posts: 18
Credit: 30970261
RAC: 1653

mikey wrote:ES-Six

mikey wrote:

ES-Six wrote:

I investigated more :

I read more HTTP debug logs of my boinc clients and I observed links and domains used :

I notices the client attempt to download files from two domains einstein-dl.syr.edu and einstein2.aei.uni-hannover.de domain. 

Do you use a VPN? If so try telling it you are in the US and see what happens, Try Canada, South America etc etc as since it works for some people you just need to make it connect to those Servers instead of the 2 that are failing for you.

Personally I don't use a VPN.

But, that's interesting, I think you think einstein@home is doing a sort of load balancing based on geographic location of the user IP si I will try using a VPN and see what happens.

ES-Six
Joined: 6 Oct 18
Posts: 18
Credit: 30970261
RAC: 1653

Richard Haselgrove wrote: I

Richard Haselgrove wrote:

I don't think is as simple as 'single server down'. I have several machines which are still downloading and processing GW tasks as normal - currently working on frequencies between 0627.20 and 0637.20 (the next group of figures after the h1_ prefix). I think that the problem tends to pounce when we've finished all the tasks for a given frequency band, and the project tries to send us the data files for a new frequency band. If they're not ready, or have gone awol - boom!

With regard to the URLs: Einstein@Home has multiple servers around the world. In Europe, I get sent

<file><br />
    <name>h1_0635.80_O3aC01Cl1In0</name><br />
    <nbytes>17582296.000000</nbytes><br />
    <max_nbytes>0.000000</max_nbytes><br />
    <md5_cksum>4f85b13047f9f20465a174337ff51311</md5_cksum><br />
    <status>1</status><br />
    <sticky/><br />
    <download_url>http://einstein2.aei.uni-hannover.de/download/8c/h1_0635.80_O3aC01Cl1In0</download_url><br />
    <download_url>http://einstein-dl.syr.edu/download/8c/h1_0635.80_O3aC01Cl1In0</download_url><br />
</file>

with the European server (hannover.de) listed first. I think that people further west - in the Americas - get directed to (?) Syracuse University (syr.edu) in preference. The data files on both servers should be kept as exact mirrors of each other.

The project administrators don't really have time to sit and read these messages. They usually rely on the forum moderators to alert them to developing problems. I had hoped that Gary Roberts might have picked up on this one: if he hasn't commented by tomorrow, I'll reach out to Bernd myself.

Ok, it makes sense, I will try with a VPN to see what happens.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.