Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen
The flag (of type list) you can use is "--least_cost_functions" Multiple algorithms can be specified, the default one is "compute_fill_first_cost_fn" which gives priority to the compute host (XenServer/ESX etc.) with more free RAM. Regards, Sateesh "This e-mail message is for the sole use of the intended recipient(s) and may contain confidential and/or privileged information. Any unauthorized review, use, disclosure, or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply e-mail and destroy all copies of the original message." From: openstack-bounces+sateesh.chodapuneedi=citrix@lists.launchpad.net [mailto:openstack-bounces+sateesh.chodapuneedi=citrix@lists.launchpad.net] On Behalf Of Razique Mahroua Sent: Thursday, November 10, 2011 4:12 PM To: Jorge Luiz Correa Cc: openstack@lists.launchpad.net Subject: Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen +1 :) Razique Mahroua razique.mahr...@gmail.com<mailto:razique.mahr...@gmail.com> [cid:image002.jpg@01CC9FD5.EC50CAF0] Le 10 nov. 2011 à 11:27, Jorge Luiz Correa a écrit : Is there a flag in nova.conf that permits we configure that? In documentation we can see that exists some algorithms used by scheduler. But, I don't know how to choose that one best fit our requirements. Thanks! :) On Wed, Nov 9, 2011 at 1:36 PM, Ed Leafe mailto:ed.le...@rackspace.com>> wrote: On Nov 9, 2011, at 7:51 AM, Razique Mahroua wrote: > I use the default scheduler, in fact, I've never tunned it really. > The hypervisors all run KVM This is where the flag is defined in nova.scheduler.least_cost.py<http://nova.scheduler.least_cost.py/>: 32 FLAGS = flags.FLAGS 33 flags.DEFINE_list('least_cost_functions', 34 ['nova.scheduler.least_cost.compute_fill_first_cost_fn'], 35 'Which cost functions the LeastCostScheduler should use.') Since the default weighting function is 'compute_fill_first_cost_fn', which, as its name suggests, chooses hosts so as to fill up one host as much as possible before selecting another, the pattern you're seeing is expected. If you change that flag to 'nova.scheduler.noop_cost_fn', you should see the hosts selected randomly. The idea is that you can create your own weighting functions that will select potential hosts in a way that best fits your needs. -- Ed Leafe ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net<mailto:openstack@lists.launchpad.net> Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp -- - MSc. Correa, J.L. ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net<mailto:openstack@lists.launchpad.net> Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp <><>___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen
Are you using Diablo or Trunk? If you're using trunk the default scheduler is MultiScheduler, which uses Chance scheduler. I think Diablo uses Chance by default? --scheduler_driver Unless you've explicitly selected the LeastCostScheduler (which only exists in Diablo now) I wouldn't worry about those settings. Did you explicitly define a scheduler to use? From: openstack-bounces+sandy.walsh=rackspace@lists.launchpad.net [openstack-bounces+sandy.walsh=rackspace@lists.launchpad.net] on behalf of Jorge Luiz Correa [corre...@gmail.com] Sent: Thursday, November 10, 2011 6:27 AM To: openstack@lists.launchpad.net Subject: Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen Is there a flag in nova.conf that permits we configure that? In documentation we can see that exists some algorithms used by scheduler. But, I don't know how to choose that one best fit our requirements. Thanks! :) On Wed, Nov 9, 2011 at 1:36 PM, Ed Leafe mailto:ed.le...@rackspace.com>> wrote: On Nov 9, 2011, at 7:51 AM, Razique Mahroua wrote: > I use the default scheduler, in fact, I've never tunned it really. > The hypervisors all run KVM This is where the flag is defined in nova.scheduler.least_cost.py<http://nova.scheduler.least_cost.py>: 32 FLAGS = flags.FLAGS 33 flags.DEFINE_list('least_cost_functions', 34 ['nova.scheduler.least_cost.compute_fill_first_cost_fn'], 35 'Which cost functions the LeastCostScheduler should use.') Since the default weighting function is 'compute_fill_first_cost_fn', which, as its name suggests, chooses hosts so as to fill up one host as much as possible before selecting another, the pattern you're seeing is expected. If you change that flag to 'nova.scheduler.noop_cost_fn', you should see the hosts selected randomly. The idea is that you can create your own weighting functions that will select potential hosts in a way that best fits your needs. -- Ed Leafe ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net<mailto:openstack@lists.launchpad.net> Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp -- - MSc. Correa, J.L. ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen
+1 :) Razique Mahrouarazique.mahr...@gmail.com Le 10 nov. 2011 à 11:27, Jorge Luiz Correa a écrit :Is there a flag in nova.conf that permits we configure that? In documentation we can see that exists some algorithms used by scheduler. But, I don't know how to choose that one best fit our requirements. Thanks!:)On Wed, Nov 9, 2011 at 1:36 PM, Ed Leafewrote: On Nov 9, 2011, at 7:51 AM, Razique Mahroua wrote: > I use the default scheduler, in fact, I've never tunned it really. > The hypervisors all run KVM This is where the flag is defined in nova.scheduler.least_cost.py: 32 FLAGS = flags.FLAGS 33 flags.DEFINE_list('least_cost_functions', 34 ['nova.scheduler.least_cost.compute_fill_first_cost_fn'], 35 'Which cost functions the LeastCostScheduler should use.') Since the default weighting function is 'compute_fill_first_cost_fn', which, as its name suggests, chooses hosts so as to fill up one host as much as possible before selecting another, the pattern you're seeing is expected. If you change that flag to 'nova.scheduler.noop_cost_fn', you should see the hosts selected randomly. The idea is that you can create your own weighting functions that will select potential hosts in a way that best fits your needs. -- Ed Leafe ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp -- - MSc. Correa, J.L. ___Mailing list: https://launchpad.net/~openstackPost to : openstack@lists.launchpad.netUnsubscribe : https://launchpad.net/~openstackMore help : https://help.launchpad.net/ListHelp___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen
Is there a flag in nova.conf that permits we configure that? In documentation we can see that exists some algorithms used by scheduler. But, I don't know how to choose that one best fit our requirements. Thanks! :) On Wed, Nov 9, 2011 at 1:36 PM, Ed Leafe wrote: > On Nov 9, 2011, at 7:51 AM, Razique Mahroua wrote: > > > I use the default scheduler, in fact, I've never tunned it really. > > The hypervisors all run KVM > >This is where the flag is defined in nova.scheduler.least_cost.py: > > 32 FLAGS = flags.FLAGS > 33 flags.DEFINE_list('least_cost_functions', > 34 ['nova.scheduler.least_cost.compute_fill_first_cost_fn'], > 35 'Which cost functions the LeastCostScheduler should use.') > >Since the default weighting function is > 'compute_fill_first_cost_fn', which, as its name suggests, chooses hosts so > as to fill up one host as much as possible before selecting another, the > pattern you're seeing is expected. If you change that flag to > 'nova.scheduler.noop_cost_fn', you should see the hosts selected randomly. > The idea is that you can create your own weighting functions that will > select potential hosts in a way that best fits your needs. > > > -- Ed Leafe > > > ___ > Mailing list: https://launchpad.net/~openstack > Post to : openstack@lists.launchpad.net > Unsubscribe : https://launchpad.net/~openstack > More help : https://help.launchpad.net/ListHelp > -- - MSc. Correa, J.L. ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen
Hi Razique, Did you synchronize clock of all the servers? I saw similar issue when clock of nova servers were not synchronized well. In this case, compute nodes are recognized as down and up (smiley) in turn repeatedly. Setting up an NTP server is a good idea, and please check using ntpq command if all nova nodes are synchronized. Regards, Masanori From: Razique Mahroua Subject: Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen Date: Wed, 9 Nov 2011 16:14:35 +0100 > Hi Akira, > I did, 4 nodes are here, up and smiling :-) > Like I said, I explicitly disable the first three nodes in order to force the > scheduler, and it worked ; the last node was choosen for the new instance I > spawned, without any issue. > > > > Le 9 nov. 2011 à 16:03, Akira Yoshiyama a écrit : > > > Hi, > > > > Did you try "nova-manage --flagfile=/etc/nova/nova.conf service list"? > > > > Regards, > > Akira Yoshiyama > > > > 2011/11/09 18:52 "Razique Mahroua" : > > Hl all, > > I've four compute-nodes, around 20 instances runing. > > I've four nodes registered to nova-scheduler (nova-manage shows them) > > but everytime I spawn a new instance, the 3rd an 4th node are never choosen > > for the instances. > > the ressources are the same on the nodes (around 24gb of ram), they are > > idle, and available. > > Can I force the scheduler to use them for an instance ? > > Thanks > > Razique - doc team - > > ___ > > Mailing list: https://launchpad.net/~openstack > > Post to : openstack@lists.launchpad.net > > Unsubscribe : https://launchpad.net/~openstack > > More help : https://help.launchpad.net/ListHelp > ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen
I would like to understand that too. When I was testing, in some cases, a 16 GB node were with no instance while a 2 GB host ran 3 or 4 instances. And, new instances were to the 2 GB node, even all the nodes 'smiling'. Thanks! On Wed, Nov 9, 2011 at 1:14 PM, Razique Mahroua wrote: > Hi Akira, > I did, 4 nodes are here, up and smiling :-) > Like I said, I explicitly disable the first three nodes in order to force > the scheduler, and it worked ; the last node was choosen for the new > instance I spawned, without any issue. > > > > Le 9 nov. 2011 à 16:03, Akira Yoshiyama a écrit : > > Hi, > > Did you try "nova-manage --flagfile=/etc/nova/nova.conf service list"? > > Regards, > Akira Yoshiyama > 2011/11/09 18:52 "Razique Mahroua" : > >> Hl all, >> I've four compute-nodes, around 20 instances runing. >> I've four nodes registered to nova-scheduler (nova-manage shows them) >> but everytime I spawn a new instance, the 3rd an 4th node are never >> choosen for the instances. >> the ressources are the same on the nodes (around 24gb of ram), they are >> idle, and available. >> Can I force the scheduler to use them for an instance ? >> Thanks >> Razique - doc team - >> ___ >> Mailing list: https://launchpad.net/~openstack >> Post to : openstack@lists.launchpad.net >> Unsubscribe : https://launchpad.net/~openstack >> More help : https://help.launchpad.net/ListHelp >> > > > ___ > Mailing list: https://launchpad.net/~openstack > Post to : openstack@lists.launchpad.net > Unsubscribe : https://launchpad.net/~openstack > More help : https://help.launchpad.net/ListHelp > > -- - MSc. Correa, J.L. ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen
On Nov 9, 2011, at 7:51 AM, Razique Mahroua wrote: > I use the default scheduler, in fact, I've never tunned it really. > The hypervisors all run KVM This is where the flag is defined in nova.scheduler.least_cost.py: 32 FLAGS = flags.FLAGS 33 flags.DEFINE_list('least_cost_functions', 34 ['nova.scheduler.least_cost.compute_fill_first_cost_fn'], 35 'Which cost functions the LeastCostScheduler should use.') Since the default weighting function is 'compute_fill_first_cost_fn', which, as its name suggests, chooses hosts so as to fill up one host as much as possible before selecting another, the pattern you're seeing is expected. If you change that flag to 'nova.scheduler.noop_cost_fn', you should see the hosts selected randomly. The idea is that you can create your own weighting functions that will select potential hosts in a way that best fits your needs. -- Ed Leafe ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen
Hi Akira, I did, 4 nodes are here, up and smiling :-) Like I said, I explicitly disable the first three nodes in order to force the scheduler, and it worked ; the last node was choosen for the new instance I spawned, without any issue. Le 9 nov. 2011 à 16:03, Akira Yoshiyama a écrit : > Hi, > > Did you try "nova-manage --flagfile=/etc/nova/nova.conf service list"? > > Regards, > Akira Yoshiyama > > 2011/11/09 18:52 "Razique Mahroua" : > Hl all, > I've four compute-nodes, around 20 instances runing. > I've four nodes registered to nova-scheduler (nova-manage shows them) > but everytime I spawn a new instance, the 3rd an 4th node are never choosen > for the instances. > the ressources are the same on the nodes (around 24gb of ram), they are idle, > and available. > Can I force the scheduler to use them for an instance ? > Thanks > Razique - doc team - > ___ > Mailing list: https://launchpad.net/~openstack > Post to : openstack@lists.launchpad.net > Unsubscribe : https://launchpad.net/~openstack > More help : https://help.launchpad.net/ListHelp ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] Four compute-node, everytime the 1st and 2nd are choosen
Hi, Did you try "nova-manage --flagfile=/etc/nova/nova.conf service list"? Regards, Akira Yoshiyama 2011/11/09 18:52 "Razique Mahroua" : > Hl all, > I've four compute-nodes, around 20 instances runing. > I've four nodes registered to nova-scheduler (nova-manage shows them) > but everytime I spawn a new instance, the 3rd an 4th node are never > choosen for the instances. > the ressources are the same on the nodes (around 24gb of ram), they are > idle, and available. > Can I force the scheduler to use them for an instance ? > Thanks > Razique - doc team - > ___ > Mailing list: https://launchpad.net/~openstack > Post to : openstack@lists.launchpad.net > Unsubscribe : https://launchpad.net/~openstack > More help : https://help.launchpad.net/ListHelp > ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
[Openstack] Four compute-node, everytime the 1st and 2nd are choosen
Hl all, I've four compute-nodes, around 20 instances runing. I've four nodes registered to nova-scheduler (nova-manage shows them) but everytime I spawn a new instance, the 3rd an 4th node are never choosen for the instances. the ressources are the same on the nodes (around 24gb of ram), they are idle, and available. Can I force the scheduler to use them for an instance ? Thanks Razique - doc team - ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp