Re: HPC cluster setup
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Juan, I'm not quite sure what you are asking. Can you give be more specific about what you are trying to do? Josh On Wednesday, April 11, 2012 8:56:55 PM jjzam...@itesm.mx wrote: How you define the configuration of the parent node machine and the subimages machines? Thanks Juan José Zamanillo -Original Message- From: Josh Thompson josh_thomp...@ncsu.edu Date: Wed, 11 Apr 2012 16:09:43 To: vcl-user@incubator.apache.org Reply-To: vcl-user@incubator.apache.org Subject: Re: HPC cluster setup - gpg control packet Al, Pick an image to be the parent. Then, edit that image's profile by going to Manage Images and clicking Edit for the image. Click the Advanced Options section to expand it. Click the Manage Subimages button. Add any images you want to be part of the cluster (note that these changes are immediate; you don't click a save button or anything). Now, when you make a reservation for the parent image, you will get the whole cluster. Subimages are not recursive - i.e. if you add a subimage that has subimages assigned to it, you only get the one subimage, not the subimage's subimages as well. Josh On Wednesday, April 11, 2012 7:39:08 PM Evelio Quiros wrote: Hello, Now that I have VCL working to provide single virtual machines, how do I setup a cluster of machines in VCL ? How do I setup the situation where a user can reserve a small cluster out of a group of resources (a head node and child compute nodes) ? Thanks, Al Quiros - -- - --- Josh Thompson VCL Developer North Carolina State University my GPG/PGP key can be found at pgp.mit.edu All electronic mail messages in connection with State business which are sent to or received by this account are subject to the NC Public Records Law and may be disclosed to third parties. -BEGIN PGP SIGNATURE- Version: GnuPG v2.0.17 (GNU/Linux) iEYEARECAAYFAk+GzAAACgkQV/LQcNdtPQPPIACfePjOHU/HnoOFMOVgf8Ug4o1b 7fkAnjihcKpEV62RhJDIs3gnhxdwP31e =BNd4 -END PGP SIGNATURE-
Re: HPC cluster setup
Currently the only cluster configuration that is done is a cluster_info file is populated on each node (parent and children) with the IP addresses and their level. something like: 123.45.67.890=parent 123.45.67.891=child 123.45.67.892=child etc This cluster_info file is either in /etc for linux or C: for windows So with that info an image creator can use that file from and init scripts and start their cluster. We have several hadoop clusters here at NCSU using this method, it's simple but seems to work well. In the future we would be better served using configuration managers like puppet, chef, bcfg2, etc. (We should put this on the roadmap.. :) ) Hope this helps. Aaron On Wed, Apr 11, 2012 at 4:56 PM, jjzam...@itesm.mx wrote: How you define the configuration of the parent node machine and the subimages machines? Thanks Juan José Zamanillo -Original Message- From: Josh Thompson josh_thomp...@ncsu.edu Date: Wed, 11 Apr 2012 16:09:43 To: vcl-user@incubator.apache.org Reply-To: vcl-user@incubator.apache.org Subject: Re: HPC cluster setup -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Al, Pick an image to be the parent. Then, edit that image's profile by going to Manage Images and clicking Edit for the image. Click the Advanced Options section to expand it. Click the Manage Subimages button. Add any images you want to be part of the cluster (note that these changes are immediate; you don't click a save button or anything). Now, when you make a reservation for the parent image, you will get the whole cluster. Subimages are not recursive - - i.e. if you add a subimage that has subimages assigned to it, you only get the one subimage, not the subimage's subimages as well. Josh On Wednesday, April 11, 2012 7:39:08 PM Evelio Quiros wrote: Hello, Now that I have VCL working to provide single virtual machines, how do I setup a cluster of machines in VCL ? How do I setup the situation where a user can reserve a small cluster out of a group of resources (a head node and child compute nodes) ? Thanks, Al Quiros - -- - --- Josh Thompson VCL Developer North Carolina State University my GPG/PGP key can be found at pgp.mit.edu All electronic mail messages in connection with State business which are sent to or received by this account are subject to the NC Public Records Law and may be disclosed to third parties. -BEGIN PGP SIGNATURE- Version: GnuPG v2.0.17 (GNU/Linux) iEYEARECAAYFAk+F5Q0ACgkQV/LQcNdtPQMZnwCfYlKTdwQmtySQV4Kb29ms/s+K neAAnArGLL4qs93a1qRssL9PibtRid5w =bhO6 -END PGP SIGNATURE- -- Aaron Peeler Program Manager Virtual Computing Lab NC State University All electronic mail messages in connection with State business which are sent to or received by this account are subject to the NC Public Records Law and may be disclosed to third parties.
Re: openstack and VCL ?
Hi Cameron, Sounds great. Our team are also interested in implementing the OpenStack Provisioning module to VCL at IBM tivioli. The stage here is also in the proof of concept and I've tried to build up the test environment now. But before I go further, I'd like to know whether you already implemented the openstack provisioning module or not. I'd like to avoid any duplicating efforts on the same work. If you already finished, please let me know. Also, I'll join the ICA CON 2012 and I'm looking forward to your presentation. We can discuss more details about your work after it. Thank you. Young From: Aaron Peeler aaron_pee...@ncsu.edu To: Cameron Mann cameron.m...@cybera.ca, Cc: vcl-user@incubator.apache.org, vcl-...@incubator.apache.org Date: 03/29/2012 01:09 PM Subject:Re: openstack and VCL ? Great. I'll create a jira issue on it to give it an initial home. You should see it come across the vcl-dev list shortly. Thanks again and looking forward to your presentation. Aaron On Thu, Mar 29, 2012 at 11:08 AM, Cameron Mann cameron.m...@cybera.ca wrote: We'd definitely be comfortable with that and should be able to do so within the next two or three weeks. Cameron On Wed, Mar 28, 2012 at 11:08 AM, Aaron Peeler aaron_pee...@ncsu.edu wrote: Hi Cameron, That's wonderful. I understand this is a POC now, but also thats a great point in development cycle to get more eyes looking at it. Would you be comfortable (if not now, soon) submitting the work under ASF for review by the VCL community? Also with that we could work toward bringing you or your lead on this effort in as a apache vcl committer. I look forward to seeing you at ICA CON 2012 Best Regards, Aaron On Wed, Mar 28, 2012 at 11:50 AM, Cameron Mann cameron.m...@cybera.ca wrote: Hi Aaron, One of the things we're working on at Cybera is an EC2 provisioning module. We've also done a lot of work with OpenStack and our intent is for the module not just to work with Amazon EC2 but any cloud that implements the EC2 API, including OpenStack. Right now it's very much in a proof of concept state, but our intent is to open source it once some of the rough edges are smoothed over. We'll also be presenting a paper at the upcoming ICA CON 2012 on our work so far. Cameron On Tue, Mar 27, 2012 at 8:28 AM, Aaron Peeler aaron_pee...@ncsu.edu wrote: Hi Folks, I think Tony from ECU asked about this before. Has anyone worked with openstack http://openstack.org/ ? I think integrating openstack provisioning into the VCL framework would benefit VCL. It could also benefit openstack and their current users. From what I understand, Openstack has a large commercial base of users and could be a way to expose VCL to more commercial users. Unless I'm wrong, currently openstack doesn't provide a self-service interface for users to VDI or clusters, nor does it provide bare-metal loads. VCL with xCAT could provide that for that community in a short time-frame. Is there anyone interested in investigating and working on an openstack provisioning module ? I don't think it will take a huge amount of work, one would need to know openstack and how to add in a VCL provisioning module. I'd be willing to assist but would need someone to take the lead on it. Thoughts? I know this is a development question, but also felt this is big enough to include folks on the vcl-user list. Best, Aaron -- Aaron Peeler Program Manager Virtual Computing Lab NC State University All electronic mail messages in connection with State business which are sent to or received by this account are subject to the NC Public Records Law and may be disclosed to third parties. -- Aaron Peeler Program Manager Virtual Computing Lab NC State University All electronic mail messages in connection with State business which are sent to or received by this account are subject to the NC Public Records Law and may be disclosed to third parties. -- Aaron Peeler Program Manager Virtual Computing Lab NC State University All electronic mail messages in connection with State business which are sent to or received by this account are subject to the NC Public Records Law and may be disclosed to third parties. inline: graycol.gif
available computers
Hi, All I found that sometimes there is a 'disconnect' between information in SQL database and actual state of 'available' computers. If VCL has a bunch of availabe computers preloaded with an image and I then update that image (new image revision), computers are not automatically reloaded with newer image revision. New reservation will have new image revision, but those available computer will still have older image revision and it would be reloaded only if a user makes reservation. At that time it will take longer to load (ie. ~10min vs 1min for preloaded available computer). Also in cases when I need to reboot ESXi hosts, all the VMs are not loaded automatically. In this case database lists those VMs as available, but they are off-line. I would like to share the small tool (a .pl script), which syncs available computers with actual state of computers (VMs). I've used peaces of code from utils.php to write it. The script creates reload requests for all available computers using the current image and latest image revision. VCLD then caries out requests and reloads computers which are either have older image revision or cannot be contacted via SSH (ie. after ESXi reboot). I plan to run the script weekly via cron on vcld server to make sure all is in sync, as well as after image updates. I've tested it and it works fine. .pl is attached. (I have hardcoded $managementnodeid value in reloadComputer(). Yours could be different) Thanks. -- Dmitri Chebotarov Virtual Computing Lab Systems Engineer, TSD - Ent Servers Messaging 223 Aquia Building, Ffx, MSN: 1B5 Phone: (703) 993-6175 Fax: (703) 993-3404 imageReload.pl Description: Binary data
Re: Thin client with VCL
Thank you for sharing the information - these are all good points. Thank you. -- Dmitri Chebotarov Virtual Computing Lab Systems Engineer, TSD - Ent Servers Messaging 223 Aquia Building, Ffx, MSN: 1B5 Phone: (703) 993-6175 Fax: (703) 993-3404 On Tuesday, April 10, 2012 at 11:26 , Henry Schaffer wrote: Michael's point about testing is excellent! :-) Other considerations include comparing prices - the price of desktop computer has decreased greatly in the past few years, narrowing the price difference from thin clients. Also consider how you might want to use the desktop/local machines. The VCL is a desktop augmentation setup - so you likely want to use the local machines for web surfing, e-mail, perhaps word processing ..., what else. That decision impacts the price of the thin client. --henry schaffer On Tue, Apr 10, 2012 at 10:59 AM, Michael Jinks mji...@uchicago.edu (mailto:mji...@uchicago.edu) wrote: Several years ago, we put thin clients (Sun Ray) in all our public computing spaces and computer-equipped classrooms. They work great for most things, and they do indeed save lots of expense and hassle. We're now in the process of going back to PC's, though. There are several reasons, but the one that might apply to other sites is remote display of graphically-intensive applications. 3D rendering is the obvious one, but there are also a few legacy (DOS-era) scientific graphing packages that don't play well with a network-connected display, and the accumulated latency during real-time graphing appears to the user as a drastic slowdown in performance. So, test all your apps thoroughly before you commit. On Tue, Apr 10, 2012 at 09:41:11AM -0400, Dmitri Chebotarov wrote: Hi All, Is anyone here is using a thin client with VCL? I.e. Dell FX100 or similar? (http://en.wikipedia.org/wiki/Thin_client). This could work well with VCL, since most of thin boxes support RDP. Interesting to see how a thin client compares to a regular PC in classroom environment. Seems like this would be a better option - less expensive, less admin overhead, more secure, and with all the benefits of VCL... Thanks. -- Dmitri Chebotarov Virtual Computing Lab Systems Engineer, TSD - Ent Servers Messaging 223 Aquia Building, Ffx, MSN: 1B5 Phone: (703) 993-6175 Fax: (703) 993-3404 -- Michael Jinks :: mji...@uchicago.edu (mailto:mji...@uchicago.edu) :: 773-469-9688 University of Chicago IT Services
Re: available computers
Cool. Nice script. Do you have an account under Jira? https://issues.apache.org/jira/browse/VCL If so would recommend to create a jira issue for 2.4 release to add this in. This way we can make sure to add that in. Also it may be worth looking at a script called healthcheck.pl in pathtovcl/vcl/bin. I think healthcheck.pl does not work too well in 2.2.1 - had some issues with the data structure object. The healthcheck.pl in the repo for the 2.3 is pretty stable, it doesn't initiate a reload, but checks the state of the machine, current image in db vs actual image loaded and updates the database accordingly. So maybe, we could combine yours and healthcheck. Aaron On Thu, Apr 12, 2012 at 12:04 PM, Dmitri Chebotarov dcheb...@gmu.edu wrote: Hi, All I found that sometimes there is a 'disconnect' between information in SQL database and actual state of 'available' computers. If VCL has a bunch of availabe computers preloaded with an image and I then update that image (new image revision), computers are not automatically reloaded with newer image revision. New reservation will have new image revision, but those available computer will still have older image revision and it would be reloaded only if a user makes reservation. At that time it will take longer to load (ie. ~10min vs 1min for preloaded available computer). Also in cases when I need to reboot ESXi hosts, all the VMs are not loaded automatically. In this case database lists those VMs as available, but they are off-line. I would like to share the small tool (a .pl script), which syncs available computers with actual state of computers (VMs). I've used peaces of code from utils.php to write it. The script creates reload requests for all available computers using the current image and latest image revision. VCLD then caries out requests and reloads computers which are either have older image revision or cannot be contacted via SSH (ie. after ESXi reboot). I plan to run the script weekly via cron on vcld server to make sure all is in sync, as well as after image updates. I've tested it and it works fine. .pl is attached. (I have hardcoded $managementnodeid value in reloadComputer(). Yours could be different) Thanks. -- Dmitri Chebotarov Virtual Computing Lab Systems Engineer, TSD - Ent Servers Messaging 223 Aquia Building, Ffx, MSN: 1B5 Phone: (703) 993-6175 Fax: (703) 993-3404 -- Aaron Peeler Program Manager Virtual Computing Lab NC State University All electronic mail messages in connection with State business which are sent to or received by this account are subject to the NC Public Records Law and may be disclosed to third parties.