I have deployed puppet on our test cluster -- 20 identically configured x86_64 and ppc64 compute nodes and their gateways. I now want to move to our production cluster with (say) 800 compute nodes and their gateways. My question is about the puppet server box itself. I am using apache+ssl with passenger and ruby 1.8.6. It was easy to install and set up apache with passenger and to get puppet running maintaining various files and services across the test cluster. The puppet server is a dual-cpu x86 box with 1GB ram. It handles the test cluster without troubles. Any thoughts about what sort of puppet-box I need to manage a cluster of 800 boxes? Comments would be very much appreciated. Cheers-- Charles -- Advanced Computing Center for Research and Education Vanderbilt University --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Puppet Users" group. To post to this group, send email to puppet-users@googlegroups.com To unsubscribe from this group, send email to puppet-users+unsubscribe@googlegroups.com For more options, visit this group at http://groups.google.com/group/puppet-users?hl=en -~----------~----~----~----~------~----~------~--~---
On 8/19/09 10:06 AM, "cnjohnson" <gm.johns276@gmail.com> wrote:> Any thoughts about what sort of puppet-box I need to manage a cluster > of 800 boxes?How often will puppet "wake up" on each client? If only once an hour normally, a decent "normal" server (say 8GB RAM, dual dual-core CPU''s) should be fine. At least that''s what we''ve found with our particular mix of configs. --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Puppet Users" group. To post to this group, send email to puppet-users@googlegroups.com To unsubscribe from this group, send email to puppet-users+unsubscribe@googlegroups.com For more options, visit this group at http://groups.google.com/group/puppet-users?hl=en -~----------~----~----~----~------~----~------~--~---
On Aug 19, 2009, at 10:06 AM, cnjohnson wrote:> > I have deployed puppet on our test cluster -- 20 identically > configured x86_64 and ppc64 compute nodes and their gateways. I now > want to move to our production cluster with (say) 800 compute nodes > and their gateways. My question is about the puppet server box itself. > I am using apache+ssl with passenger and ruby 1.8.6. It was easy to > install and set up apache with passenger and to get puppet running > maintaining various files and services across the test cluster. The > puppet server is a dual-cpu x86 box with 1GB ram. It handles the test > cluster without troubles. > > Any thoughts about what sort of puppet-box I need to manage a cluster > of 800 boxes? >Hi Charles, It depends upon a number of factors, mostly your Puppet manifests, if you are using storeconfig and how much file handling you are doing. If you are using store config, you may want to think about moving the mysql (or posgresql) onto another server and/or additional CPUs. If the puppetmaster is x86_64 based, I would recommend 4+ GB based upon your node count and they way ruby scales on x86_64 arch. You should at minimum be on 0.24.8 and ideally 0.25 once it''s released. Hope this helps.. -L -- Larry Ludwig Reductive Labs --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Puppet Users" group. To post to this group, send email to puppet-users@googlegroups.com To unsubscribe from this group, send email to puppet-users+unsubscribe@googlegroups.com For more options, visit this group at http://groups.google.com/group/puppet-users?hl=en -~----------~----~----~----~------~----~------~--~---
On Thu, Aug 20, 2009 at 9:02 PM, Larry Ludwig <larry@reductivelabs.com>wrote:> > If the puppetmaster is x86_64 based, I would recommend 4+ GB based > upon your node count and they way ruby scales on x86_64 arch. >I install 32bit distributions on my 64bit hardware, as a single passenger/mongrel doesn''t (usually) pass the 3GB size, it doesn''t matter and i think behaves a bit better.> > You should at minimum be on 0.24.8 and ideally 0.25 once it''s released. >+1 for 0.24.8, we have way less problems with this version. Ohad --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Puppet Users" group. To post to this group, send email to puppet-users@googlegroups.com To unsubscribe from this group, send email to puppet-users+unsubscribe@googlegroups.com For more options, visit this group at http://groups.google.com/group/puppet-users?hl=en -~----------~----~----~----~------~----~------~--~---
Thanks to everyone for his/her replies. I appreciate it. We will be adding RAM to the box. We are using puppet 0.24.8. We are not using storeconfig at the moment, and if we were we would be using our postgres server for that. I appreciate that pointer. The size of the files we serve are, typically, tiny: sudoers, hosts, hosts.equiv, and eight or ten others. They do not change often, but when they do, they need to be synched across the cluster. Our big bug-a-boo is seeing that new nodes are properly provisioned. The other issue we face is keeping a set of services running on nodes. autofs is notorious for us. Again, thanks for the replies. Cheers-- Charles -- Advanced Computing Center for Research and Education Vanderbilt University --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Puppet Users" group. To post to this group, send email to puppet-users@googlegroups.com To unsubscribe from this group, send email to puppet-users+unsubscribe@googlegroups.com For more options, visit this group at http://groups.google.com/group/puppet-users?hl=en -~----------~----~----~----~------~----~------~--~---
On Fri, Aug 21, 2009 at 2:22 AM, Charles Johnson <gm.johns276@gmail.com>wrote:> > The size of the files we serve are, typically, tiny: sudoers, hosts, > hosts.equiv, and eight or ten others. They do not change often, but when > they do, they need to be synched across the cluster. Our big bug-a-boo is > seeing that new nodes are properly provisioned. The other issue we face is > keeping a set of services running on nodes. autofs is notorious for us. >I would recommend you not to allow autofs to restart, but rather force a reload (if your OS supports it). (it would be similar to pulling the plug out of the machine :)) Just as a side note, do you have any software that manages your cluster ? (e.g. LSF/ Sungrid etc?) We have a distributed cluster with a couple of thousands of nodes and we are very grateful for having puppet.... Ohad --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Puppet Users" group. To post to this group, send email to puppet-users@googlegroups.com To unsubscribe from this group, send email to puppet-users+unsubscribe@googlegroups.com For more options, visit this group at http://groups.google.com/group/puppet-users?hl=en -~----------~----~----~----~------~----~------~--~---
On Thu, Aug 20, 2009 at 9:26 PM, Ohad Levy <ohadlevy@gmail.com> wrote:> > I would recommend you not to allow autofs to restart, but rather force a > reload (if your OS supports it). > (it would be similar to pulling the plug out of the machine :)) > > Just as a side note, do you have any software that manages your cluster ? > (e.g. LSF/ Sungrid etc?) > We have a distributed cluster with a couple of thousands of nodes and we > are very grateful for having puppet.... > > Ohad >We use moab for scheduling and torque for resource management. We also use IBM''s gpfs for a networked file system. We have to support a connection to Sun''s networked file system --BlueArc-- for which we use autofs as the automounter. We currently use nagios for monitoring. All of our x86 nodes are CentOS 5.3 while the PowerPC''s are RHEL. Cheers-- Charles --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Puppet Users" group. To post to this group, send email to puppet-users@googlegroups.com To unsubscribe from this group, send email to puppet-users+unsubscribe@googlegroups.com For more options, visit this group at http://groups.google.com/group/puppet-users?hl=en -~----------~----~----~----~------~----~------~--~---