Thread 'Running Headless - Problem Connecting to the Projects'

Message boards : Questions and problems : Running Headless - Problem Connecting to the Projects
Message board moderation

To post messages, you must log in.

AuthorMessage
Adak

Send message
Joined: 16 Oct 08
Posts: 7
United States
Message 41138 - Posted: 13 Nov 2011, 2:59:21 UTC

I run some headless PC's (no keyboard,monitor,etc), that crunch 24/7 but have only occasional net access. No more than 1 hour per day.

Problem is, when I connect them, the Windows rigs won't d/l their results. They are put onto a "project backoff" count down timer, which delays it for hours at a time (up to 12 hours in some cases). That doesn't work.

How can I get BOINC Manager to:

Upload finished work and d/l new work when the network is connected, instead of this "project backoff", BS?

(I have a Linux box that also runs headless, and it handles it beautifully, so I know it's within the realm of the possible.)

OS: Windows 7 x64
Not sure what the BOINC Manager version is, but it's less than a year old.

Thanks a lot! ;)
ID: 41138 · Report as offensive
ProfileGundolf Jahn

Send message
Joined: 20 Dec 07
Posts: 1069
Germany
Message 41140 - Posted: 13 Nov 2011, 7:23:02 UTC - in response to Message 41138.  

Is the client on those PCs allowed network access when not connected? That would explain the backoffs.

Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
ID: 41140 · Report as offensive
Adak

Send message
Joined: 16 Oct 08
Posts: 7
United States
Message 41144 - Posted: 13 Nov 2011, 11:35:03 UTC - in response to Message 41139.  
Last modified: 13 Nov 2011, 11:37:38 UTC

I'm not sure I understand you're situation. It sounds like you have several computers but only 1 network cable so you move the cable from 1 computer to another to give each one some connection time. Is that right?

Exactly. They were on a router, and the router somehow lost it's brains, and I can't find them! ;)

I would think the BOINC developers would also want a smarter program. Trying to connect when the network access is not present, and then refusing to connect when the network is present...

Seems quite backwards. And, the Linux version of BOINC (an older version on Ubuntu 9.04), handles it just fine. Just the newer Windows 7 version is backwards.

No access whatsoever, except when the network cable gets plugged into the modem
ID: 41144 · Report as offensive
ProfileJord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15541
Netherlands
Message 41145 - Posted: 13 Nov 2011, 11:44:53 UTC - in response to Message 41144.  

I would think the BOINC developers would also want a smarter program. Trying to connect when the network access is not present, and then refusing to connect when the network is present...

You can set the Network Activity to Suspended before you remove the network cable and put it back on "based on preferences" or Always available when you connect it again. Then you will only have the back-off when the project does not have internet access (e.g. when their servers are down).

BOINC cannot determine between the computer having no physical internet connection, the network cable being removed, the network card being broken, your ISP being down, the project servers not being up etc. etc. as they all have the same symptoms: There's no answer from over there.

Now, this can also be because the project servers are under too much stress, a deluge of computers trying to reach them, in a sense doing a directional denial of service attack. That's where the back-off comes in. For the "just in case there is a connection, but we're part of the DDoS attack, let's stop trying" cases.

When in back-off, you can do an Update on the project. When that doesn't work, go Advanced->Do network communication. Either of these will shake the back-off loose. And after you had all communications, suspend your network activity.
ID: 41145 · Report as offensive
Adak

Send message
Joined: 16 Oct 08
Posts: 7
United States
Message 41148 - Posted: 13 Nov 2011, 19:48:05 UTC - in response to Message 41145.  

I would think the BOINC developers would also want a smarter program. Trying to connect when the network access is not present, and then refusing to connect when the network is present...

You can set the Network Activity to Suspended before you remove the network cable and put it back on "based on preferences" or Always available when you connect it again. Then you will only have the back-off when the project does not have internet access (e.g. when their servers are down).

BOINC cannot determine between the computer having no physical internet connection, the network cable being removed, the network card being broken, your ISP being down, the project servers not being up etc. etc. as they all have the same symptoms: There's no answer from over there.

Now, this can also be because the project servers are under too much stress, a deluge of computers trying to reach them, in a sense doing a directional denial of service attack. That's where the back-off comes in. For the "just in case there is a connection, but we're part of the DDoS attack, let's stop trying" cases.

When in back-off, you can do an Update on the project. When that doesn't work, go Advanced->Do network communication. Either of these will shake the back-off loose. And after you had all communications, suspend your network activity.


Is that REALLY your argument?

1) I can't make changes to BOINC - these PC's run *HEADLESS* - no mouse, no monitor, no keyboard - just a PC, crunching.

2) BOINC can know when the network is not connected - for whatever reason - from the system. Easy as pie. Some versions have this ability, as I mentioned previously. The most popular versions of BOINC however (Windows), DO NOT have this ability, and should have it.

3) There is no reason at all to "back off" when there is no internet whatsoever.

4) There is no reason to continue "backing off" when the internet connection is re-established.

Backing off is a good function to use *when the PC has an internet connection*, and could easily tie up a server. It's a bad function to use when the PC is not connected to the internet. It needs to STOP the back off, when the internet connection is renewed.

No, an Update does not reset the backoff timer. You have to close down BOINC, or even reset the computer, in some cases. On the older BOINC version for Linux, it resets itself, automatically, as soon as you have an internet connection. That's what BOINC should be doing (and isn't), in the newer Windows versions (this is 6.12.23 (x64).

ID: 41148 · Report as offensive
ProfileJord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15541
Netherlands
Message 41149 - Posted: 13 Nov 2011, 20:26:01 UTC - in response to Message 41148.  

1) I can't make changes to BOINC - these PC's run *HEADLESS* - no mouse, no monitor, no keyboard - just a PC, crunching.

Sure you can, please read http://boinc.berkeley.edu/wiki/Controlling_BOINC_remotely and then explain why you can't do that.

2) BOINC can know when the network is not connected - for whatever reason - from the system. Easy as pie. Some versions have this ability, as I mentioned previously. The most popular versions of BOINC however (Windows), DO NOT have this ability, and should have it.

The client BOINC is compiled from one source code for all platforms. So when one platform can do it (Linux) all others (Windows, OS X) should be able to do so as well. But since BOINC is a platform neutral program, it doesn't support OS specific things such as checking if the network is there. That's built into the client, through Curl, but then Curl has to know about all the different forms of no connection possible.

3) There is no reason at all to "back off" when there is no internet whatsoever.

4) There is no reason to continue "backing off" when the internet connection is re-established.

{snip}

All your other Windows programs will either go on some sort of retry timer when there's no internet connection, retrying every so often if the situation has changed, or throw an error immediately and not try ever again.

Why isn't BOINC allowed to do this then, in a friendly manner? Please do remember that you are not the only person using BOINC and any of the projects your BOINC is attached to. Do you really want a lot of computers to DDoS a project's servers if they have had a problem and they have just come back? Is it that important?

The project back-off is reset when doing an Update.
I had a back-off running on the Einstein project for my ATI card. Einstein doesn't have work for the ATI card, so it backs off automatically to its 24 hour retry cycle.
<rsc_backoff_time>
<name>ATI</name>
<value>76800.000000</value>
</rsc_backoff_time>
<rsc_backoff_interval>
<name>ATI</name>
<value>1321239125.891645</value>
</rsc_backoff_interval>
[quote]

I then did an Update on the project:
[quote]<rsc_backoff_time>
<name>ATI</name>
<value>0.000000</value>
</rsc_backoff_time>
<rsc_backoff_interval>
<name>ATI</name>
<value>0.000000</value>
</rsc_backoff_interval>


Now, of course, you do not have to believe me, as what do I know? ;-)

ID: 41149 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5121
United Kingdom
Message 41152 - Posted: 13 Nov 2011, 23:33:31 UTC

And back to the subject of BOINC.

BOINC can, and does, try to distinguish between "can't connect to server" and "can't connect to internet".

The former is likely to be a server problem, and requires anti-DDOS measures when service is resumed. The latter is likely to be a local problem, so doesn't.
ID: 41152 · Report as offensive
Adak

Send message
Joined: 16 Oct 08
Posts: 7
United States
Message 41179 - Posted: 16 Nov 2011, 5:39:28 UTC

Well Jord, I've tried it over 40 times, on two different systems, and you're asking me to either believe my eyes, or you.

You can't win that argument.

Update resets the backoff timer only if you stop BOINC, and then re-start it, or re-boot the system. In addition, a restart of BOINC will not reset all backoff timers, if there are multiple "groups" of backed off work units, the last "group" keeps their back off timer as is, and the first group only, is reset.

This is on Windows 7, with 6.12.26

If the programmers prefer it, how about a button we can click on to reset the back off timer?

Programmers develop a long list of excuses why something can't be done:

1) It's just not possible (I love the sheer majestic sweep of this one!)
2) No other program in that OS does it.
3) It will put <something> in grave danger
4) We don't have the <manpower,resources,money,time,etc.> for this

etc.

Please don't bother passing them around to me. I remember when BOINC couldn't find it's ass with both hands, in a well-lit room. Now look how far you've come with BOINC?

Don't stop, and don't accept excuses - help BOINC go forward. I'll be working with Folding@Home for awhile; get a few more million points over there.










ID: 41179 · Report as offensive

Message boards : Questions and problems : Running Headless - Problem Connecting to the Projects

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.