[sisuite-users] System Configurator hang-ups
I think I know what the problem is, but I'm not sure how to compensate for it. Here's the symptom: So, I created USB autoinstall disks yesterday and imaged a classroom of Dell 170Ls. With visions of them all working this morning, I left. This morning was not so rosy. I had not re-installed system configurator with -prefix=/USR on that particular golden client yet, so it bombed at that point. Having already faced this problem before, the first thing I did was reinstall system configurator with the correct configuration switch, and then I ran a test autoinstall on a single workstation, which went quite quickly, and where it had before bombed out from a lack of system configurator it spat out this: end_request: I/O error, dev fd0, sector 0 end_request: I/O error, dev fd0, sector 0 Buffer I/O error on device fd0, logical block 0 end_request: I/O error, dev fd0, sector 0 Buffer I/O error on device fd0, logical block 0 Buffer I/O error on device fd0, logical block 0 etc. Perplexed, I attempted to autoinstall on another test machine, a Dell 210L, and it took the time to image itself, and flew right past the part where these errors kept occuring, and finished the imaging process with no problem, while the 170L kept spitting out these error lines. My hunch is that it's spitting these out because the golden client has a CD drive and the machines to be imaged do not, and it just won't give up on it's hunt. Might there be a way to have system configurator give up sooner? or just not worry about the CD drive at all? Ah, I just checked again, and it seems that eventually, over a half an hour later, it does give up and get on with it. Any ideas? I hope you're all having a great day! Simón - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
Re: [sisuite-users] System Configurator hang-ups
Simon, FD0 is a floppy disk or sometimes can be mapped to a USB key. -Drew XLHost.com -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Simon Ruiz Sent: Tuesday, March 27, 2007 8:49 AM To: sisuite-users Subject: [sisuite-users] System Configurator hang-ups I think I know what the problem is, but I'm not sure how to compensate for it. Here's the symptom: So, I created USB autoinstall disks yesterday and imaged a classroom of Dell 170Ls. With visions of them all working this morning, I left. This morning was not so rosy. I had not re-installed system configurator with -prefix=/USR on that particular golden client yet, so it bombed at that point. Having already faced this problem before, the first thing I did was reinstall system configurator with the correct configuration switch, and then I ran a test autoinstall on a single workstation, which went quite quickly, and where it had before bombed out from a lack of system configurator it spat out this: end_request: I/O error, dev fd0, sector 0 end_request: I/O error, dev fd0, sector 0 Buffer I/O error on device fd0, logical block 0 end_request: I/O error, dev fd0, sector 0 Buffer I/O error on device fd0, logical block 0 Buffer I/O error on device fd0, logical block 0 etc. Perplexed, I attempted to autoinstall on another test machine, a Dell 210L, and it took the time to image itself, and flew right past the part where these errors kept occuring, and finished the imaging process with no problem, while the 170L kept spitting out these error lines. My hunch is that it's spitting these out because the golden client has a CD drive and the machines to be imaged do not, and it just won't give up on it's hunt. Might there be a way to have system configurator give up sooner? or just not worry about the CD drive at all? Ah, I just checked again, and it seems that eventually, over a half an hour later, it does give up and get on with it. Any ideas? I hope you're all having a great day! Simón - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
Re: [sisuite-users] System Configurator hang-ups
There may be a setting in the bios which lets you select how the USB key is mounted (either floppy or virtual hard disk) I wouldn't know why you're having this problem really, it seems like you could boot all of these boxes over the LAN and install simultaneously. That's what we do here. Thanks, -Drew -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Simon Ruiz Sent: Tuesday, March 27, 2007 9:27 AM To: sisuite-users@lists.sourceforge.net Subject: Re: [sisuite-users] System Configurator hang-ups Well, I did use a usb key to start the auto-install process, but pulled it out as soon as the initrd was loaded. This same proceedure on the 210Ls does not cause the same problem. In fact, I notice it seems to map fd0 and hd0 even though the only two drives it has mounted are the hard disk and a CD drive. The 170Ls do not have CD drives. This is why I was under the suspicion that the CD drive has something to do with it. It would sort of ruin the niceness of being able to pull the USB key out and move on to the next machine if I had to leave the USB key plugged in during the whole autoinstall process... Thanks for your reply! Simón From: [EMAIL PROTECTED] on behalf of Drew Weaver Sent: Tue 3/27/2007 9:11 AM To: sisuite-users@lists.sourceforge.net Subject: Re: [sisuite-users] System Configurator hang-ups Simon, FD0 is a floppy disk or sometimes can be mapped to a USB key. -Drew XLHost.com - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
Re: [sisuite-users] System Configurator hang-ups
Well, I don't control the network here, I can't make any changes to the DHCP server, so network booting is not possible AFAIK, and since the 170Ls don't have CD drives, USB keys are my only choice. Hmmm, looking in the BIOS of the 170Ls, I'm remembering I had to turn on the floppy drive in the BIOS so they could recognize the USB floppy drives we used back when we imaged these computers using Norton Ghost. Sounds promising, I'm going to turn that off and try again... Success! Thanks for the pointer! Simón From: [EMAIL PROTECTED] on behalf of Drew Weaver Sent: Tue 3/27/2007 9:33 AM To: sisuite-users@lists.sourceforge.net Subject: Re: [sisuite-users] System Configurator hang-ups There may be a setting in the bios which lets you select how the USB key is mounted (either floppy or virtual hard disk) I wouldn't know why you're having this problem really, it seems like you could boot all of these boxes over the LAN and install simultaneously. That's what we do here. Thanks, -Drew - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
Re: [sisuite-users] Cool!
We've been using SIS for a while now (in fact, I'm sure we need to update) and are very happy with it. So I feel a little funny asking this question since it seems like a no-brainer and we should already know the answer. Nonetheless, the fact remains that we have not figured out how to have the nodes reboot themselves after they have been successfully imaged. Instead, we must go in and see that they are saying I've been done for xxx seconds/minutes. , and then reboot or power-cycle them. I know that we tried at one point to overcome this problem and eventually just gave up. Yet, it seems so fundamental that I know we are doing something wrong. What is the secret to getting the nodes to just reboot themselves when they are done so that they come up ready to go after being imaged without manual intervention? Thanks, Charlie Taylor UF HPC Center - Charles A. Taylor, Ph.D Senior HPC Systems Engineer UF HPC Center University of Florida Gainesville, FL 32611 (352) 392-4036 - On Mar 16, 2007, at 11:45 AM, Andrea Righi wrote: Simon Ruiz wrote: So, the classroom imaged alright overnight. It seems like the server ran out of memory and had to kill a few processes during the course of the imaging process, but every workstation seemed to image fine. The beeping classroom was rather amusing when I came in. Might this locking up of the server be alleviated by popping in some more memory? (it's only got 512 right now, I believe, it's just exactly the same type of hardware as the workstations, really) or will it simply eat up as much memory as it gets and still lock up? You should try to use the BitTorrent trasnport. It strongly reduces the load on the image server and you should see a huge improvement in the total time of deployment... obviously only if you'll have to reimage your clients... Other than that, let me see...Ummm...ok, yeah, so when I do an si_updateclient, it seems to wipe out every change I make during the autoconfig script that first time, EXCEPT for the hostname, which keeps the workstation from realizing it needs to run the autoconfig script again. I can change it to test the contents of /etc/hosts/ instead of the $HOSTNAME variable, though I suppose I should really figure out how all the systemimager scripts work and see if maybe they would help me accomplish my goals in an easier way. Also, is there any way to trigger an autoinstall on the workstation and pass it the imagename? trying to do si_updateclient with both --autoinstall and --image doesn't seem to work... I've introduced, only in the trunk for now, the new command si_pushoverrides. This implements a new different approach to distribute files to the clients. Theoretically we can define 3 different levels of files distribution, ordered from the most dangerous to the safest: 1) imaging the nodes (a complete re-installation from scratch) 2) reset the nodes to a default known installation (via si_updateclient) 3) distribute files (typically configuration files) to customize a default installation (via si_pushoverrides) Obviously with the first approach you need to completely stop the production in your machines during the deployment. Typically you should use this method only at the beginning of your cluster / servers production or when you strongly do changes on the distributions that you're using in your clients. The second approach can be used to reset the image to a known state. The command si_updateclient, in fact, performs a re-sync of the whole client filesystem (except some customizable exclusions, see /etc/systemimager/updateclient.local.exclude). In this case you must evaluate if this process is dangerous enough for your applications, because you can lose some of your customizations and you could change something that you don't expect (remember that the update is done in the whole image.. I repeat: it is done in the whole image! :-) In general is the sysadmin that should evaluate if this operation can be dangerous enough to stop production or not... But, at the moment, there is not a totally safe way to distribute files to the clients in SystemImager. This is the reason because I wanted to add the si_pushoverrides command. With this command you can use the overrides to push only a set of files from the image server to your clients (using rsync over SSH for the backend). In this way the whole image is not touched, but only the files you put in the defined overrides can be overwritten. Using a configuration file you can also define the topology of your clients. This can be very useful for clusters that have different subsets of nodes, like compute nodes, frontend nodes, I/O servers, etc based on the same image. I should say that si_pushoverrides is under development, but I'm
Re: [sisuite-users] Cool!
Hey, Not too difficult to implement once you know about it, but took me some documentation trudging to figure out. When GETTING the image, using si_getimage, just add the switch --post-install reboot at the end, and the autoinstall script will be updated accordingly. HTH, Simón From: [EMAIL PROTECTED] on behalf of Charles A. Taylor Sent: Tue 3/27/2007 10:22 AM To: sisuite-users@lists.sourceforge.net Cc: Craig Prescott Subject: Re: [sisuite-users] Cool! We've been using SIS for a while now (in fact, I'm sure we need to update) and are very happy with it. So I feel a little funny asking this question since it seems like a no-brainer and we should already know the answer. Nonetheless, the fact remains that we have not figured out how to have the nodes reboot themselves after they have been successfully imaged. Instead, we must go in and see that they are saying I've been done for xxx seconds/minutes. , and then reboot or power-cycle them. I know that we tried at one point to overcome this problem and eventually just gave up. Yet, it seems so fundamental that I know we are doing something wrong. What is the secret to getting the nodes to just reboot themselves when they are done so that they come up ready to go after being imaged without manual intervention? Thanks, Charlie Taylor UF HPC Center - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
[sisuite-users] Bittorrent Not Quite Working...
I've been following http://wiki.systemimager.org/index.php/HOWTO_Implement _a_simple_BitTorrent_and_USB_stick_imaging_scheme but when the clients boot from the USB key, they seem to get stuck when trying to grab an image through bittorrent, and remind me to turn the bittorrent server on. I didn't install from rpms, but I compiled and installed 3.8.0 server_all and client_all on my server and golden clients respectively which, I've been under the impression, should install all the same stuff that would be installed by installing all the various packages. This may be the problem, if I'm wrong about this. I've already been preparing clients and getting the images and deploying them via rsync, so I know I'm doing that part right. I've configured /etc/systemimager/bittorrent.conf with the line BT_IMAGES=210,170,1300 so that all the images are allowed to be bittorrented, and added /etc/init.d/systemimager-server-bittorrent start to my /etc/rc.local right after the /etc/init.d/systemimager-server-rsyncd start line. A ps aux | grep systemimager shows me the I have /usr/bin/rsync --daemon --config=/etc/systemimager/rsyncd.conf and /usr/bin/python /usr/bin bttrack --port 6969 --dfile /tmp/dstate --logfile /var/log/systemimager/bittorrent-tracker.log running... Alright, I think I'm going to try not invoking it on boot I'll comment that out and try to run it by hand to see what messages it gives... ERROR: cannot find a valid binary to make torrents failed. So I search through the synaptic package manager and find libtorrent7, but installing it doesn't change anything, so I'm going to try installing libtorrent7-dev and recompiling systemimager to see if that might make a difference... Well, I've learned that any settings changes that are made to /etc/systemimager/bittorrent.conf need to be re-made after re-installing systemimager. However, the same ERROR: cannot find a valid binary to make torrents. message pops up when I try to start the bittorrent server, so I'm guessing the libtorrent7 and libtorrent7-dev aren't of any use... Any ideas? I hope this finds you all doing well! Simón - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
Re: [sisuite-users] Bittorrent Not Quite Working...
Simon Ruiz wrote: ERROR: cannot find a valid binary to make torrents failed. So I search through the synaptic package manager and find libtorrent7, but installing it doesn't change anything, so I'm going to try installing libtorrent7-dev and recompiling systemimager to see if that might make a difference... I think you may be looking for a package named bittorrent instead of libtorrent. Also, you shouldnt have to recompile systemimager, i dont think... - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
Re: [sisuite-users] Bittorrent Not Quite Working...
Andrew, Yeah, that's the first one I checked, and it was already installed, I believe by default. Version 3.4.2-6ubuntu3 (edgy). I thought maybe it was just a client, it is described as contains the tools which are used for console-only downloading. (Not to mention, it doesn't seem to be what systemimager is looking for.) So I grasped at straws with the library and the dev library + a recompile. Simón From: [EMAIL PROTECTED] on behalf of Andrew M. Lyons Sent: Tue 3/27/2007 11:48 AM To: sisuite-users@lists.sourceforge.net Subject: Re: [sisuite-users] Bittorrent Not Quite Working... Simon Ruiz wrote: ERROR: cannot find a valid binary to make torrents failed. So I search through the synaptic package manager and find libtorrent7, but installing it doesn't change anything, so I'm going to try installing libtorrent7-dev and recompiling systemimager to see if that might make a difference... I think you may be looking for a package named bittorrent instead of libtorrent. Also, you shouldnt have to recompile systemimager, i dont think... - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
[sisuite-users] Ack! Too many!
So, I've learned that until I get SystemImager working with bittorrent, I can't image more than a classroom (32 workstations) at a time. I just attempted to image two classrooms at a time, and even after running a sudo /etc/init.d/gdm stop and closing down the GUI on my server to clear out as much memory as I can, it keeps complaining: [17189932.728000] Out of memory: Kill process 4431 (rsync) score 5980 and children. [17189932.728000] Out of memory: Killed process 5247 (rsync). And disconnecting from one of the workstations that are supposed to be imaging right now. I'm going to have to show up early tomorrow morning and start the disconnected ones again. Lesson learned. Hope you're all having a great day! Simón - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
[sisuite-users] Installation is incomplete after transferring image from image server.
Hello, I have hard time with installing Image to a client. I used systemimager v3.8.0 rpm and PXE. During installation I get blank screen when 50% of my image was transfered, and it doesn't reboot when it had enough time to finish the installation. I checked on the rsycnd logs and find out scripts/imaging_complete file is missing. how can i get the scripts/imaging_complete file? what do i have to do to coplete the installation successfully? After Installation process(I waited enough time to finish up installation) I reboot the client manually and it was booted successfully but without swap partition which was included in the golden client image. Any ideas for what could be wrong? === [EMAIL PROTECTED] systemimager]# vi /var/log/systemimager/rsyncd 2007/03/21 19:19:24 [31648] rsync on boot/x86_64/standard/boel_binaries. tar.gz from dell2.pspace.org (192.168.11.102) 2007/03/21 10:19:24 [31648] wrote 5815618 bytes read 122 bytes total size 5814783 2007/03/21 19:19:26 [31649] rsync on scripts/ from dell2.pspace.org (192.168.11.102) 2007/03/21 10:19:26 [31649] wrote 89386 bytes read 251 bytes total size 88526 2007/03/21 19:20:57 [31656] rsync on dell2_image from dell2.pspace.org (192.168.11.102) 2007/03/21 19:22:54 [31656] wrote 9240258 bytes read 79 bytes total size 8824404182 2007/03/21 19:22:55 [31660] rsync on dell2_image/ from dell2.pspace.org (192.168.11.102) 2007/03/21 19:33:46 [31660] wrote 8680820509 bytes read 6872601 bytes total size 8824404182 2007/03/21 19:33:53 [31668] rsync on overrides/dell2_image/ from dell2.pspace.org (192.168.11.102) 2007/03/21 10:33:53 [31668] wrote 56 bytes read 102 bytes total size 0 2007/03/21 19:33:57 [31669] rsync on scripts/imaging_complete from dell2.pspace.org (192.168.11.102) 2007/03/21 10:33:57 [31669] rsync: link_stat imaging_complete (in scripts) failed: No such file or directory (2) 2007/03/21 10:33:57 [31669] rsync error: some files could not be transferred (code 23) at main.c(442) - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
Re: [sisuite-users] [Systemimager] Bittorrent Not Quite Working...
On Mar 27, 2007, at 7:02 PM, Simon Ruiz wrote: Andrew, Yeah, that's the first one I checked, and it was already installed, I believe by default. Version 3.4.2-6ubuntu3 (edgy). I thought maybe it was just a client, it is described as contains the tools which are used for console-only downloading. (Not to mention, it doesn't seem to be what systemimager is looking for.) So I grasped at straws with the library and the dev library + a recompile. Simón You must install version 4.0 or higher, the 3.4.2 version does not have the software to make torrents. Regards -- Bas van der Vlies [EMAIL PROTECTED] - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ sisuite-users mailing list sisuite-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-users
