Pros: Better performance. Cons: If you have a block storage associated to compute node, and the compute node dies, you won’t be able to live migrate your VMs to the other compute nodes.
From: Andalabs [mailto:mahardika.gil...@andalabs.com] Sent: Thursday, September 05, 2013 4:45 PM To: Gangur, Hrushikesh (HP Converged Cloud - R&D - Sunnyvale) Cc: Openstack Milis Subject: Re: [Openstack] Question About Multinode and Swift What it's pros and con? I need to know because i am in process preparing some server ro deploy. On 5 Sep 2013, at 11:19 PM, "Gangur, Hrushikesh (HP Converged Cloud - R&D - Sunnyvale)" <hrushikesh.gan...@hp.com<mailto:hrushikesh.gan...@hp.com>> wrote: Yes, it could be a SAN - A block storage attached to compute node. But, it has its pros and cons. From: Mahardhika [mailto:mahardika.gil...@andalabs.com] Sent: Wednesday, September 04, 2013 11:46 PM To: Gangur, Hrushikesh (HP Converged Cloud - R&D - Sunnyvale) Cc: Openstack Milis Subject: Re: [Openstack] Question About Multinode and Swift Yeah i thought of that way, could it be SAN that attach to /var/lib/nova/instance? so in this way swift is just for database / object thing . thanks so much. On 9/5/2013 1:36 PM, Gangur, Hrushikesh (HP Converged Cloud - R&D - Sunnyvale) wrote: Yes, the compute node’s sizing must be done in terms of CPU, Memory and Disk based on the flavor being selected. If you want to host 50 vms that have flavor specs 1 vcpu, 2 gb RAM, 10 gb root disk each, you need a compute node with 500 GB instance repository, 72 GB Memory, 16 core CPU. Instead of local repository, you may like to mount /var/lib/nova/instances as an NFS. This way it would be compute node independent and you easily migrate VM instances across compute nodes (in case of failure). Regards~hrushi From: Mahardhika [mailto:mahardika.gil...@andalabs.com] Sent: Wednesday, September 04, 2013 11:28 PM To: Gangur, Hrushikesh (HP Converged Cloud - R&D - Sunnyvale) Cc: Openstack Milis Subject: Re: [Openstack] Question About Multinode and Swift So, in this case we can't separate them to another server? so thats mean we need large hardisk for compute node, is that right? On 9/5/2013 12:48 PM, Gangur, Hrushikesh (HP Converged Cloud - R&D - Sunnyvale) wrote: vm instances' root and ephemeral disk data are stored in compute node's /var/lib/nova/instances. for storing user data i.e persistent, you must use cinder's block storage or swifts object store. Cheers ~hrushi On Sep 4, 2013, at 10:19 PM, "Mahardhika" <mahardika.gil...@andalabs.com><mailto:mahardika.gil...@andalabs.com> wrote: Dear all, i have some question around my head. 1. i have compute node and running KVM hypervisor in there, controller node and network node. question is, in where data that instance run store? is it on compute node? 2. if i want to store all data in separate node let say using swift (swift node), can this happen? 3. what is flavor used to store data? i mean when i create instance with selected flavor(ex:10GB), where that space that would be create? in swift node? 4. So if this can be done, in my compute node is just run KVM, and no data store that. right? Thanks before. -- Regards, Mahardhika Gilang _______________________________________________ Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack Post to : openstack@lists.openstack.org<mailto:openstack@lists.openstack.org> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack -- Regards, Mahardhika Gilang
_______________________________________________ Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack Post to : openstack@lists.openstack.org Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
