Re: Long term reservation
Aaron, If I may suggest, we should start a page on the wiki with hardware configurations people have successfully deployed on. This is a very daunting area for someone who is contemplating an installation. There is already a few hints on the wiki about NC State's hardware (but perhaps a bit more detail would be useful). It would be good if more people listed their diverse configurations in detail. Right now I have an installation on a modest desktop system (mostly to cement my understanding after last summer's bootcamp). I can put that up. I will also be doing a real deployment soon and can put that up too as I have completed it. Mark On Fri, Apr 27, 2012 at 8:42 AM, Aaron Peeler fapee...@ncsu.edu wrote: Hi Emir, I'll try to answer, but hopefully Andy will chime in to confirm. In the upcoming release, the copying vmdk for long-term reservations has been fixed. It's using snapshots to achieve. Resulting in faster boot time. On the vms per host question. This is very good question. So far your 100 vms per host is the highest I've heard about. As your aware, the number of vms and end-user performance is going to depend on the underlying hardware (host memCPU, network, and storage). It would be good as a community for us to share hardware recommendations on what is working well at their own site. We have a mix of hardware at NCSU, I'll write up some details and send that out in a separate thread soon. Aaron On Thu, Apr 26, 2012 at 3:56 PM, Emir Imamagic eimam...@srce.hr wrote: Hello, We've noticed that in case of long term reservations VCL copies the virtual disk of image on datastore. In case of images with many subimages (20) we experienced problems with this copying. VCL would initiate multiple vmkfstools commands, ESXi server would get overloaded and start killing vmkfstools processes (messages were indicating lack of memory). Is there any way to bypass this behavior? Is this copying really needed? Is it possibly to switch it off in a clean manner? Another question is - how many VMs can vcld handle per a single VM host (VMware ESXi 4.1)? On our setup we managed to start 100 VMs on a single VMware host and it was still working fine. VM host has 24 cores and 256GB RAM. Thanks in advance -- Emir Imamagic SRCE - University of Zagreb University Computing Centre, www.srce.unizg.hr emir.imama...@srce.hr, tel: +385 1 616 5809, fax: +385 1 616 5559 -- 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. -- Mark Gardner --
Re: Long term reservation
Hi Mark, Great suggestion, it's a great way to share what works(for both small test beds to larger scale deployments). Would you like to start it off? I think a basic example is fine to start with. To all, The confluence wiki is open, one just needs to create an account (if you don't already have on on the apache confluence wiki.). Aaron On Fri, Apr 27, 2012 at 9:55 AM, Mark Gardner m...@vt.edu wrote: Aaron, If I may suggest, we should start a page on the wiki with hardware configurations people have successfully deployed on. This is a very daunting area for someone who is contemplating an installation. There is already a few hints on the wiki about NC State's hardware (but perhaps a bit more detail would be useful). It would be good if more people listed their diverse configurations in detail. Right now I have an installation on a modest desktop system (mostly to cement my understanding after last summer's bootcamp). I can put that up. I will also be doing a real deployment soon and can put that up too as I have completed it. Mark On Fri, Apr 27, 2012 at 8:42 AM, Aaron Peeler fapee...@ncsu.edu wrote: Hi Emir, I'll try to answer, but hopefully Andy will chime in to confirm. In the upcoming release, the copying vmdk for long-term reservations has been fixed. It's using snapshots to achieve. Resulting in faster boot time. On the vms per host question. This is very good question. So far your 100 vms per host is the highest I've heard about. As your aware, the number of vms and end-user performance is going to depend on the underlying hardware (host memCPU, network, and storage). It would be good as a community for us to share hardware recommendations on what is working well at their own site. We have a mix of hardware at NCSU, I'll write up some details and send that out in a separate thread soon. Aaron On Thu, Apr 26, 2012 at 3:56 PM, Emir Imamagic eimam...@srce.hr wrote: Hello, We've noticed that in case of long term reservations VCL copies the virtual disk of image on datastore. In case of images with many subimages (20) we experienced problems with this copying. VCL would initiate multiple vmkfstools commands, ESXi server would get overloaded and start killing vmkfstools processes (messages were indicating lack of memory). Is there any way to bypass this behavior? Is this copying really needed? Is it possibly to switch it off in a clean manner? Another question is - how many VMs can vcld handle per a single VM host (VMware ESXi 4.1)? On our setup we managed to start 100 VMs on a single VMware host and it was still working fine. VM host has 24 cores and 256GB RAM. Thanks in advance -- Emir Imamagic SRCE - University of Zagreb University Computing Centre, www.srce.unizg.hr emir.imama...@srce.hr, tel: +385 1 616 5809, fax: +385 1 616 5559 -- 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. -- Mark Gardner -- -- 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: Long term reservation
A general comment from someone who has watched the process and seen some testing and failures. This is a bottleneck type problem - the major resources are 1) RAM - used for the VM manager/host and then for the images (which can vary in size) 2) processors/cores- i.e. computer horsepower 3) disk throughput (for loading images) Whichever runs out first will give the limit - which can be a limit on number loaded (RAM), adequate performance for the ones loaded (horsepower) or time to load them (disk throughput). We and others in the VCL community have run into all of these at times, and there have been a number of neat solutions - e.g. remove unneeded software from the image - which was rather important back in the days of small RAMs. So I suggest that along with hardware configurations to be discussed on the wiki, that also some more information be given as to the above. --henry
Re: Apache LIBCLOUD driver contribution to interact with Apache VCL
I.e in English you say computer, in Mexico whe say computadora and in Spain they say ordenador. Other example in English you say file in Mexico we say archivo and in Spain they say fichero Juan José Zamanillo -Original Message- From: jjzam...@itesm.mx Date: Thu, 26 Apr 2012 19:24:29 To: vcl-user@incubator.apache.org Reply-To: vcl-user@incubator.apache.org Subject: Re: Apache LIBCLOUD driver contribution to interact with Apache VCL It is spanish but The terminology is very different in south america, central america, mexico and spain Juan José Zamanillo -Original Message- From: Oscar Tejada otej...@gmail.com Date: Thu, 26 Apr 2012 14:36:03 To: vcl-user@incubator.apache.org Reply-To: vcl-user@incubator.apache.org Subject: Re: Apache LIBCLOUD driver contribution to interact with Apache VCL Yes there are important differences. I think an additional would be beneficial for the community Still willing to contribute and more than happy to provide you this additional translation On Apr 26, 2012 1:49 PM, Josh Thompson josh_thomp...@ncsu.edu wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Oscar, Juan Zamanillo from Tech Monterrey contributed Spanish translation files. You can download the files from subversion if you want to have a look: https://svn.apache.org/repos/asf/incubator/vcl/trunk/web/js/nls/es_MX/messages.js https://svn.apache.org/repos/asf/incubator/vcl/trunk/web/locale/po_files/es_MX/vcl.po Any idea if there are any notable differences between Spanish in Venezuela and in Mexico that would make it worth having separate translation files? Thanks, Josh On Wednesday, April 25, 2012 7:53:14 PM Oscar Tejada wrote: Greetings, This message is to let you know we are still willing to contribute to the Spanish translation of the VCL system Please advise further instructions to proceed Best regards Oscar Tejada (unitec college - Venezuela) On Apr 25, 2012 9:21 AM, Aaron Peeler fapee...@ncsu.edu wrote: Hi Folks, Just wanted to mention something cool I recently found out about. An NCSU student has contributed to Apache Libcloud community a VCL driver to interact with VCL instances. https://issues.apache.org/jira/browse/LIBCLOUD-180 From what I understand he is working on a libcloud VCL provision module, I'll try to found out more and encourage to contribute to Apache VCL. Cheers, 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. - -- - --- 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+ZkZAACgkQV/LQcNdtPQMCIACcCiuJwiBS7ZeHDIvzj3fz1s07 VVUAmwUr6Sxj+M6ZvgBo3frmEh/mgO75 =8qsC -END PGP SIGNATURE-