| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Machines reregister (typically) each time they are turned on. |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 12
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have added 19 WinXP machines at a SeniorNet Learning Center to a previous Linux box that has no problems.
Install details: The BOINC client was installed (as a daemon) on 1 WinXP machine, but NOT registered. The client was installed in its own 2Gb partition (G:). That machine was 'cloned' to 19 others. Each of the 19 was separately 'registered' on the WCG, and then DeepFreeze (basic edition) was enabled on the C: partition. Problem(s): These machines are used for about 8 hours/day, Mon-Fri, and off the rest of the time. Each time that they are turned on, they appear to re-register as a new install. I have also noticed that the WCG sometimes gives 2 machines the same ID. Some work is being reported as 'valid', but most is lost. NOTE: due to other considerations, 'cloning' is the ONLY viable option for installing the client - I was told in previous communications that the information in the 'registry' (protected by DeepFreeze AFTER registration) would not be a problem. That only the data in the 'install' directory was dynamic. Secondary problem - The WCG needs to provide a method to 'clean up' the number of machines registered. This user now shows 103+ machines (we only have 20) and going up at the rate of ~15 per work day. Ron Springer Lab Maintenance Coord. SeniorNet Learning Center - Willows San Jose, CA |
||
|
|
schepers
Advanced Cruncher Canada Joined: Oct 11, 2006 Post Count: 85 Status: Offline Project Badges: 
|
The BOINC client was installed (as a daemon) on 1 WinXP machine, but NOT registered. The client was installed in its own 2Gb partition (G:). That machine was 'cloned' to 19 others. Each of the 19 was separately 'registered' on the WCG, and then DeepFreeze (basic edition) was enabled on the C: partition. Deep Freeze Basic has no Thawspace, so you created G partitions on all the machines? Did you did remember to create the Deep Freeze installation file by specifying that G drive is not to be protected? There's nothing BOINC-related on C at all? Problem(s): These machines are used for about 8 hours/day, Mon-Fri, and off the rest of the time. Each time that they are turned on, they appear to re-register as a new install. This seems to indicate that G is being protected by Deep Freeze, and it shouldn't be. I have also noticed that the WCG sometimes gives 2 machines the same ID. Don't know about this one. NOTE: ...I was told in previous communications that the information in the 'registry' (protected by DeepFreeze AFTER registration) would not be a problem. That only the data in the 'install' directory was dynamic. This is true. The BOINC registry locations only holds manager specifics (windowing details, screen saver options), nothing else. The G partition for you holds everything else. One test... boot a machine and record some file modified dates in the BOINC folder, files that should change like client_state.xml, stdoutdae.txt. Then, after some time (hours) shut it down and reboot again. Check the dates and see if they are still the same as you first recorded. If so, Deep Freeze is still protecting them. Secondary problem - The WCG needs to provide a method to 'clean up' the number of machines registered. This user now shows 103+ machines (we only have 20) and going up at the rate of ~15 per work day. This is on the todo list, as it is becoming a much-requested website feature. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'll double check the 'thaw' state this afternoon, but I'm relatively certain that G is not frozen.
The two machines with identical ID's was noticed when only 1 of the 2 would show up at the same time on the 'My Grid > Device Manager' list. When I checked the 'html' for the 'Device Name' entries, the ID's were the same. The Linux box was registered last Nov. but I had to wait for a break in classes before I could set up the class machines (19) - I should note that when I registered the 19, I was not given the option to create a new 'account'. My only guess is that the WCG is looking at the IP address - these machines are all connected with NAT to a single IP. Is this assumption correct??? RonS |
||
|
|
schepers
Advanced Cruncher Canada Joined: Oct 11, 2006 Post Count: 85 Status: Offline Project Badges:
|
I'll double check the 'thaw' state this afternoon, but I'm relatively certain that G is not frozen. I'd be surprised. Check your Deep Freeze Administrator, and make sure that only C is being protected. I should note that when I registered the 19, I was not given the option to create a new 'account'. If you are installing BOINC, you never get the option to create a new account. You must already have a login to the WCG site. It's the UD client that asks about a new login or an existing login when you install. My only guess is that the WCG is looking at the IP address - these machines are all connected with NAT to a single IP. Is this assumption correct??? No, I doubt it. If it used the IP number, those at home who are connected directly to DSL/CABLE routers would have the IP change regularly, screwing things up for BOINC. IP numbers are much to dynamic to rely on. I'm running two machines behind a NAT router right now without troubles but they have different IP's and names. BOINC appears to use the machine name as the identifier, and then assigns it an ID number. This could mean that a machine with the same name could cause some problems. (UD uses a different method to assign an ID). Are you sure that all the PC's have different names? |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
SNLCSJwillows-VM,
----------------------------------------If you are doing an install like you are doing then you need to make sure that everything in the BOINC install directory (and lower) is protected from being overwritten each time DeepFreeze does a restore. It sounds like you have tried to install BOINC on a G partion but what I see in the database looks like the BOINC install is being overwritten periodically (sometimes multiple times in a day). The way that BOINC assigns host ids is that it checks two items each time that it contacts the servers to make sure that this is a unique computer - you can see these fields in your client_state.xml file in the BOINC install directory. In the <project> section: 1) <hostid> - this is the unique id for the computer in the system 2) <rpc_seqno> - this is the number of times that the computer has communicated with the server (the client increases this each time When the computer sends a request, the server checks these two fields to make sure the rpc_seqno is greater then the value stored on the server. If it is then the value from the client is stored and activity occurs as normal. If it is equal or lower, then this means that some other computer has communicated with the server with this id. The server then registers this host as a new device attached to the grid and sends that information down to the client. The only time that you will see the rpc_seqno be equal or lower then the value stored on the server is if you copy the boinc install and then restore it later or restore it on different machines. Becuase this is exactly the behavior we are seeing, I think that your BOINC installation locations are being overwritten every time that deep freeze restores the machine. I think that Schepers has the right idea. Another reason that I think this is what is happening is becuase your machines have downloaded over 200 workunits but only returned 20 of them. This would occur if the agent started running, requested and downlodaed work, but before it could complete the result the install location was wiped out and returned to a 'clean install' state. It would then request more work (and be identified as a different computer) and repeat the process. Let us know what the adminstrator says. thanks, Kevin [Edit 1 times, last edit by knreed at Jan 26, 2007 3:12:59 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
O.K. - DeepFreeze was/is the problem
With the 'standard' edition of DeepFreeze, you only get to set partition state at install time and I only freeze the C partition. You cannot view the freeze/thaw states after install  However, it seems that any partitions added with DeepFreeze installed are automatically given the 'freeze' state - at least that appears to be what happened. During the 8 week session of our classes, I only have limited access to the machines. I fixed?? 11 of the 19 machines yesterday and will finish on Mon. or Thu. next week - now I just would like to see the number of 'registered' machines get set back to normal  Thanks for all comments about other items, I like to understand the why's. RonS |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I forgot to answer the 'machine name' question.
They are definitely all different names, the router will not allow the same name twice - big flags RonS |
||
|
|
schepers
Advanced Cruncher Canada Joined: Oct 11, 2006 Post Count: 85 Status: Offline Project Badges:
|
I forgot to answer the 'machine name' question. They are definitely all different names, the router will not allow the same name twice - big flags I didn't know this name conflict on NAT routers existed until today, when my laptop and my daughters computer were both on... had a name conflict. Didn't realize I had named them both the same already some time ago. |
||
|
|
schepers
Advanced Cruncher Canada Joined: Oct 11, 2006 Post Count: 85 Status: Offline Project Badges:
|
O.K. - DeepFreeze was/is the problem With the 'standard' edition of DeepFreeze, you only get to set partition state at install time and I only freeze the C partition. You cannot view the freeze/thaw states after install I wonder if I can download a trial version of Deep Freeze Basic? I would like to see what differences there are between the Pro and Enterprise versions (I've used both). It would help me complete my documentation of running BOINC and UD under Deep Freeze. Thanks for all comments about other items, I like to understand the why's. Just understanding how Deep Freeze works takes time. Each version appears to protect and install a little differently. Heck, I'm still working through documenting all I'm finding out about BOINC in our environment. I've been running it since early December (almost 2 months) and I still have topic points to complete. My UD writeup was nowhere near as complex. |
||
|
|
schepers
Advanced Cruncher Canada Joined: Oct 11, 2006 Post Count: 85 Status: Offline Project Badges:
|
I've downloaded and tried the Deep Freeze Basic v6.1 installer. I trust that this is similar to whatever version you are using (possibly 5.x). I'm not used to using such a simple version- no administrator package to create a customized workstation installer. The world of Pro and Enterprise is so much more configurable and friendly.
When you ran the installer, you only had the C drive/partition showing, you could only toggle the protection for that drive, right? It appears that all the other drives are set to be frozen even though you see no checkbox for them. When you create a new partition, it is automatically frozen and in order to thaw it, you need to uninstall and re-install, unchecking the G drive (BOINC). Even when Pro and Enterprise and installed, you can't permanently change what drives are always to be frozen or thawed without uninstalling and re-installing with a new-configured installer. I've just started using Enterprise v6.1, so my playing field has changed again. I will be documenting more things about BOINC and Deep Freeze in the coming weeks, especially the Basic limitations (and side-effects) you've helped identify. |
||
|
|
|