| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Does running WCG on cloned PCs makes sense? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 19
|
|
| Author |
|
|
jochrupp
Advanced Cruncher Germany Joined: Nov 17, 2004 Post Count: 58 Status: Offline Project Badges: 
|
Hi all,
----------------------------------------I'm a school network admin and I'm thinking about installing WCG on our 120 school PCs. The question is whether this makes sense, since the PCs are cloned via an imaging system which resets them at each start to an defined state. All changes are lost so this means also all unfinished workunits would be lost. I think, whether it makes sense depends on the average uptime of the PCs which I estimate is at about 5 hrs. But it's also possible, that pupils turn them off after 90 mins of the lesson... What pevails: the opportunity to have about 120 more clients crunching or the handicap that in some (many?) cases there will be unfinished workunits? Thanks for your help and opinions! Kind regards, Jochen  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
What cloning product do you use?
There are various ways to make this work. |
||
|
|
jochrupp
Advanced Cruncher Germany Joined: Nov 17, 2004 Post Count: 58 Status: Offline Project Badges:
|
At the moment we're using Rembo2/mySHN4 from SBE. Maybe we'll change to a new opensource product (LINBO) in summer.
----------------------------------------Both of them boot the PCs via PXE, they get an IP/hostname via DHCP and after the OS image is synced/restored either from second HDD-partition or from the server, a patch is applied to set IP and hostname. All previously installed or uninstalled software or data is lost. One possibility could be to "install" the WCG programm or save the data on a third partition on the HDD that isn't synced, this data won't be lost. Thanks, Jochen |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Our generic advice is here: http://wcg.wikia.com/wiki/Mass_install_(disk_image)
However, this advice is tailored for Windows. Are you imaging the PCs with Windows? Using a third partition is possible - we will have to discuss it further. The Rembo2/mySHN4 documentation is all in German, so I may be of limited help. |
||
|
|
jochrupp
Advanced Cruncher Germany Joined: Nov 17, 2004 Post Count: 58 Status: Offline Project Badges:
|
Thanks, I'll try the installation on a third partition if possible.
----------------------------------------But what do you think about that: What prevails: the opportunity to have about 120 more clients crunching or the handicap that in some (many?) cases there will be unfinished workunits? Thanks, Jochen |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Only the data directory needs to be on a third partition (you could even use a network drive, but you will need separate directories for each computer).
As to your question of which prevails: there is not a single answer to this question. It will depend on exactly how fast your computers are, and how they are used. It also depends on how you manage them - for example, you can chose World Community Grid projects with the shortest tasks, and/or the least variation in task size. Also, future developments may improve the situation. Right now, WCG tasks are not sized ideally for this situation. But in the future, it may be possible to chose to have smaller (shorter) tasks. |
||
|
|
jochrupp
Advanced Cruncher Germany Joined: Nov 17, 2004 Post Count: 58 Status: Offline Project Badges:
|
ok, after some time has passed I'm planning to implement the mass-rollout of WCG on the network during the next weeks.
----------------------------------------The program directory will be integrated in the image, so that WCG starts every time a PC is started. At startup each machine gets its own IP and hostname via DHCP, so it's unique. When the application directory is reset to the installation state after each startup, this should be no problem, right!? In the cc_config, I'll include: <report_results_immediately>1</report_results_immediately> The data directory could be located on our NAS and as I understood, every machine needs its own data directory there!? Some more questions on that: 1. Can I use variables during the installation like: Data directory: W:\WCG\%COMPUTERNAME% ? 2. The data directory should be connected automatically when XP starts, I'll try a logon-Script with: NET USE W: \\IP_OF_NAS\WCG /PERSISTENT:YES 3. When will the data directories be populated by the client for the first time? a. During the installation process or b. when WCG starts the first time after the post-installation-reboot? I'm asking since I'm doing the installation just on one PC and then I'll clone all the other machines so the installation-process won't run at all the other machines. So if it's b. then it should be fine and each machine populates its data directory the first time it starts up. Any thoughts are appreciated! Cheers, Jochen [Edit 4 times, last edit by jochrupp at Apr 19, 2011 10:09:01 PM] |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
The data directory could be located on our NAS and as I understood, every machine needs its own data directory there!? Yes, each machine must have a unique data-directory. 1. Can I use variables during the installation like: Data directory: W:\WCG\%COMPUTERNAME% ? 2. The data directory should be connected automatically when XP starts, I'll try a logon-Script with: NET USE W: \\IP_OF_NAS\WCG /PERSISTENT:YES Can't answer on either of these. 3. When will the data directories be populated by the client for the first time? a. During the installation process or b. when WCG starts the first time after the post-installation-reboot? I'm asking since I'm doing the installation just on one PC and then I'll clone all the other machines so the installation-process won't run at all the other machines. So if it's b. then it should be fine and each machine populates its data directory the first time it starts up. The data-directory and sub-directories will be created on client-installation, but will only include a couple dummy-files. Most files will be created on running BOINC. But, while most files will be created on running BOINC, you should include the "important" files beforehand, if not you'll manually need to attach to WCG on each and every computer... The "important" file is called account_www.worldcommunitygrid.org.xml and this file among other things includes your account-key, or your weak account-key. It's also highly recommended to include two other files, gui_rpc_auth.cfg and remote_hosts.cfg, these two files is for remotely monitoring/controlling BOINC. The 1st. one is a password of your choise (can't be blank password), and the 2nd. includes one or more lines, on each line either a name or ip-address on a computer you wants to allow the client to be remotely controlled/monitored by. For <report_results_immediately> and other config-choises you'll need to include cc_config.xml. Lastly, if you needs to use a proxy with BOINC, you'll also need another file. But, unfortunately, while account-info, remote controlling and so on goes in separate files that can easily be added/edited/removed, the proxy-info goes in client_state.xml and this file also includes all info about tasks currently on computer, and will therefore be unique for each computer... So, if you needs to set proxy-info, you should do: 1: Install BOINC on a single computer. 2: DO NOT attach any projects. 3: Setup the neccessary proxy-info. 4: Exit BOINC, and use the client_state.xml you've now got to be included on all the other computers. Just to sum-up, a "clean" BOINC-data-directory for mass-installations will likely include: a: One or more account_*.xml, for all BOINC-project(s) you've attached to. If only WCG, you'll have only one account-file. b: gui_rpc_auth.cfg c: remote_hosts.cfg d: cc_config.xml e: client_state.xml, but only if needs to specify proxy-info. f: global_prefs_override.xml, if needs more specific preferences than settable on web-page, example different settings for individual days.  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." [Edit 2 times, last edit by Ingleside at Apr 19, 2011 11:14:30 PM] |
||
|
|
jochrupp
Advanced Cruncher Germany Joined: Nov 17, 2004 Post Count: 58 Status: Offline Project Badges:
|
Thanks for your answer.
----------------------------------------The data-directory and sub-directories will be created on client-installation, but will only include a couple dummy-files. Most files will be created on running BOINC. But, while most files will be created on running BOINC, you should include the "important" files beforehand, if not you'll manually need to attach to WCG on each and every computer... The "important" file is called account_www.worldcommunitygrid.org.xml and this file among other things includes your account-key, or your weak account-key. Ok, I'll provide these files in all data directories before the clients do start the first time. For <report_results_immediately> and other config-choises you'll need to include cc_config.xml. I Know. Lastly, if you needs to use a proxy with BOINC, you'll also need another file. But, unfortunately, while account-info, remote controlling and so on goes in separate files that can easily be added/edited/removed, the proxy-info goes in client_state.xml and this file also includes all info about tasks currently on computer, and will therefore be unique for each computer... So, if you needs to set proxy-info, you should do: 1: Install BOINC on a single computer. 2: DO NOT attach any projects. 3: Setup the neccessary proxy-info. 4: Exit BOINC, and use the client_state.xml you've now got to be included on all the other computers. I don't need these settings since our proxy is in transparent mode. Just to sum-up, a "clean" BOINC-data-directory for mass-installations will likely include: a: One or more account_*.xml, for all BOINC-project(s) you've attached to. If only WCG, you'll have only one account-file. b: gui_rpc_auth.cfg c: remote_hosts.cfg d: cc_config.xml e: client_state.xml, but only if needs to specify proxy-info. f: global_prefs_override.xml, if needs more specific preferences than settable on web-page, example different settings for individual days. Thanks! I'll check this out and let you know. |
||
|
|
jochrupp
Advanced Cruncher Germany Joined: Nov 17, 2004 Post Count: 58 Status: Offline Project Badges:
|
Again, "some" time passed and I've tested it last week. Unfortunately, it isn't possible to set the Data-Dirextory like this:
----------------------------------------\\NAS_IP\share\%COMPUTERNAME% Seems like %COMPUTERNAME% is the problem, \\NAS_IP\share\pc1 works. But if i use a hardcoded pc-name for the data-directory, this one goes into the image and all the other pc use the same directory. Any other idea? Thanks in advance an happy easter, Jochen |
||
|
|
|