Re: [ovirt-users] SLA : RAM scheduling
On 11 Jun 2014, at 14:31, noc wrote: On 26-5-2014 16:22, Gilad Chaplik wrote: Hi Nathanaël, happy to assist :) hope it will work in first run: 1) install the proxy and ovirtsdk. 2) put attached file in the right place (according to docs: .../plugins), make sure to edit the file with your ovirt's ip, user@domain and PW. 3) restart proxy service. 3) use config tool to configure ovirt-engine: * ExternalSchedulerServiceURL=http://ip:18781/ * ExternalSchedulerEnabled=true 4) restart ovirt-engine service. 5) under configure-cluster_policy see that weight function memory_even_distribution was added (should be in manage policy units or /sth- you will see it in the main dialog as well). 6) clone/copy currernt cluster's used cluster policy (probably none - prefer it to have no balancing modules to avoid conflicts), name it 'your_name' and attach memory_even_distribution weight (you can leave it as the only weight module in weight section to avoid configuring factors). 7) replace cluster's cluster policy with newly created one. try it out and let me know how goes :-) Ok, progress of some sort :-) I added the weight function to the cluster and when I replace my dns name with localhost in ExternalSchedulerServiceURL then engine.log shows that it can contact the scheduler. I expected a rebalance but nothing happened. Stopping and starting a VM does provoke a reaction, an error :-( From the scheduler.log I see that engine contacts it and pushes some information, the log also shows that some information is returned and then there is a big error message in the log of engine. xmlrpc is infamous about not being able to handle numbers like 9223372010239819775 Thanks, michal Joop engine.logovirt-scheduler-proxy.log___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
Michal Skrivanek wrote: On 11 Jun 2014, at 14:31, noc wrote: On 26-5-2014 16:22, Gilad Chaplik wrote: Hi Nathanaël, happy to assist :) hope it will work in first run: 1) install the proxy and ovirtsdk. 2) put attached file in the right place (according to docs: .../plugins), make sure to edit the file with your ovirt's ip, user@domain and PW. 3) restart proxy service. 3) use config tool to configure ovirt-engine: * ExternalSchedulerServiceURL=http://ip:18781/ * ExternalSchedulerEnabled=true 4) restart ovirt-engine service. 5) under configure-cluster_policy see that weight function memory_even_distribution was added (should be in manage policy units or /sth- you will see it in the main dialog as well). 6) clone/copy currernt cluster's used cluster policy (probably none - prefer it to have no balancing modules to avoid conflicts), name it 'your_name' and attach memory_even_distribution weight (you can leave it as the only weight module in weight section to avoid configuring factors). 7) replace cluster's cluster policy with newly created one. try it out and let me know how goes :-) Ok, progress of some sort :-) I added the weight function to the cluster and when I replace my dns name with localhost in ExternalSchedulerServiceURL then engine.log shows that it can contact the scheduler. I expected a rebalance but nothing happened. Stopping and starting a VM does provoke a reaction, an error :-( From the scheduler.log I see that engine contacts it and pushes some information, the log also shows that some information is returned and then there is a big error message in the log of engine. xmlrpc is infamous about not being able to handle numbers like 9223372010239819775 Then oVirt shouldn't either use that kind of numbers or should not use xmlrpc. Sorry but thats a non-answer and doesn't help anybody. Howto solve this problem. Do you need a BZ? Joop ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
Well, I know this is not that helpful but afaik there is work done to replace xml-rpc communication with json based communication. Am 13.06.2014 10:57, schrieb Joop: Then oVirt shouldn't either use that kind of numbers or should not use xmlrpc. Sorry but thats a non-answer and doesn't help anybody. Howto solve this problem. Do you need a BZ? Joop -- Mit freundlichen Grüßen / Regards Sven Kieske Systemadministrator Mittwald CM Service GmbH Co. KG Königsberger Straße 6 32339 Espelkamp T: +49-5772-293-100 F: +49-5772-293-333 https://www.mittwald.de Geschäftsführer: Robert Meyer St.Nr.: 331/5721/1033, USt-IdNr.: DE814773217, HRA 6640, AG Bad Oeynhausen Komplementärin: Robert Meyer Verwaltungs GmbH, HRB 13260, AG Bad Oeynhausen ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
On 13 Jun 2014, at 11:04, Sven Kieske wrote: Well, I know this is not that helpful but afaik there is work done to replace xml-rpc communication with json based communication. @Sven: yeah, and one of the (many) reasons why we want it. But that's vdsm. Scheduler is a different package/app. Typical workaround is to use a string type to pass integer numbers (or use smaller numbers:) Am 13.06.2014 10:57, schrieb Joop: Then oVirt shouldn't either use that kind of numbers or should not use xmlrpc. Sorry but thats a non-answer and doesn't help anybody. @Joop: what's with the attitude? I have nothing to do with the scheduler and I have no idea what is it trying or supposed to return…I'm just pointing out the error is due to the large whatever number being returned; in hope of someone with knowledge of the scheduler will take it from there. I'd tend to agree XMLRPC is not the best thing to use; and I think it's important to keep reminding the common pitfalls of it. Howto solve this problem. Do you need a BZ? patches welcome, BZ always helps. However I'd wait for someone more familiar with the code to answer. Thanks, michal Joop -- Mit freundlichen Grüßen / Regards Sven Kieske Systemadministrator Mittwald CM Service GmbH Co. KG Königsberger Straße 6 32339 Espelkamp T: +49-5772-293-100 F: +49-5772-293-333 https://www.mittwald.de Geschäftsführer: Robert Meyer St.Nr.: 331/5721/1033, USt-IdNr.: DE814773217, HRA 6640, AG Bad Oeynhausen Komplementärin: Robert Meyer Verwaltungs GmbH, HRB 13260, AG Bad Oeynhausen ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
Michal Skrivanek wrote: On 13 Jun 2014, at 11:04, Sven Kieske wrote: Well, I know this is not that helpful but afaik there is work done to replace xml-rpc communication with json based communication. @Sven: yeah, and one of the (many) reasons why we want it. But that's vdsm. Scheduler is a different package/app. Typical workaround is to use a string type to pass integer numbers (or use smaller numbers:) Am 13.06.2014 10:57, schrieb Joop: Then oVirt shouldn't either use that kind of numbers or should not use xmlrpc. Sorry but thats a non-answer and doesn't help anybody. @Joop: what's with the attitude? I have nothing to do with the scheduler and I have no idea what is it trying or supposed to return…I'm just pointing out the error is due to the large whatever number being returned; in hope of someone with knowledge of the scheduler will take it from there. I'd tend to agree XMLRPC is not the best thing to use; and I think it's important to keep reminding the common pitfalls of it. Sorry I'm only trying to help someone which got the code and it doesn't seem to work and I just figured out what is not working and replied that back to the list for everyone to see and comment on. With the now supplied reason I understand why it doesn't work but don't understand why the code is supplied in this form. Looking at the output again I don't get why the numbers are that big. The servers only have 32G and now way you use calc will it come close to the numbers seen :-( Joop ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
On 28-5-2014 17:37, Gilad Chaplik wrote: make sure to edit the file with your ovirt's ip, user@domain and PW. My engine API can't be reached in http, so there is some work to do this with https. Here is what I did : according to /usr/lib/python2.6/site-packages/ovirtsdk/api.py, I add insecure=True to the chain connection = API(url='http*s*://host:port', username='user@domain', password=''*insecure='True'*) Maybe it is not enough and it would be useful to add validate_cert_chain=False... Martin? 3) restart proxy service. 3) use config tool to configure ovirt-engine: * ExternalSchedulerServiceURL=http://ip:18781/ The scheduler proxy listens to localhost:18781, none of ips that can be filled here will be reached on that port. you config ovirt engine with the ip of the proxy, or I'm missing sth. ovirt communicates with the proxy and not the other way around. I'm following this and found out that ovirt-scheduler-proxy only listens on localhost so you'll need to adjust accordingly. Joop ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
On 26-5-2014 16:22, Gilad Chaplik wrote: Hi Nathanaël, happy to assist :) hope it will work in first run: 1) install the proxy and ovirtsdk. 2) put attached file in the right place (according to docs: .../plugins), make sure to edit the file with your ovirt's ip, user@domain and PW. 3) restart proxy service. 3) use config tool to configure ovirt-engine: * ExternalSchedulerServiceURL=http://ip:18781/ * ExternalSchedulerEnabled=true 4) restart ovirt-engine service. 5) under configure-cluster_policy see that weight function memory_even_distribution was added (should be in manage policy units or /sth- you will see it in the main dialog as well). 6) clone/copy currernt cluster's used cluster policy (probably none - prefer it to have no balancing modules to avoid conflicts), name it 'your_name' and attach memory_even_distribution weight (you can leave it as the only weight module in weight section to avoid configuring factors). 7) replace cluster's cluster policy with newly created one. try it out and let me know how goes :-) Ok, progress of some sort :-) I added the weight function to the cluster and when I replace my dns name with localhost in ExternalSchedulerServiceURL then engine.log shows that it can contact the scheduler. I expected a rebalance but nothing happened. Stopping and starting a VM does provoke a reaction, an error :-( From the scheduler.log I see that engine contacts it and pushes some information, the log also shows that some information is returned and then there is a big error message in the log of engine. Joop 2014-06-11 14:19:03,647 INFO [org.ovirt.engine.core.vdsbroker.gluster.GlusterVolumesListVDSCommand] (DefaultQuartzScheduler_Worker-79) FINISH, GlusterVolumesListVDSCommand, return: {955b86a9-10b4-463b-8555-3c321bd72f5c=org.ovirt.engine.core.common.businessentities.gluster.GlusterVolumeEntity@37074874, 0bf7869b-873b-4c28-9c09-d31100f4e12a=org.ovirt.engine.core.common.businessentities.gluster.GlusterVolumeEntity@99f2d266, 248aa48d-6aa5-4b21-867e-994265a3f145=org.ovirt.engine.core.common.businessentities.gluster.GlusterVolumeEntity@12074ee1}, log id: 6dca802f 2014-06-11 14:19:08,755 INFO [org.ovirt.engine.core.vdsbroker.gluster.GlusterVolumesListVDSCommand] (DefaultQuartzScheduler_Worker-9) [5d65b296] START, GlusterVolumesListVDSCommand(HostName = st02, HostId = 5077a01d-7273-4d58-92ee-c3315b0a973e), log id: 3af08062 2014-06-11 14:19:08,870 INFO [org.ovirt.engine.core.vdsbroker.gluster.GlusterVolumesListVDSCommand] (DefaultQuartzScheduler_Worker-9) [5d65b296] FINISH, GlusterVolumesListVDSCommand, return: {955b86a9-10b4-463b-8555-3c321bd72f5c=org.ovirt.engine.core.common.businessentities.gluster.GlusterVolumeEntity@a301dc77, 0bf7869b-873b-4c28-9c09-d31100f4e12a=org.ovirt.engine.core.common.businessentities.gluster.GlusterVolumeEntity@2cf62fb2, 248aa48d-6aa5-4b21-867e-994265a3f145=org.ovirt.engine.core.common.businessentities.gluster.GlusterVolumeEntity@b2cabf68}, log id: 3af08062 2014-06-11 14:19:13,977 INFO [org.ovirt.engine.core.vdsbroker.gluster.GlusterVolumesListVDSCommand] (DefaultQuartzScheduler_Worker-19) START, GlusterVolumesListVDSCommand(HostName = st02, HostId = 5077a01d-7273-4d58-92ee-c3315b0a973e), log id: 32bca729 2014-06-11 14:19:14,093 INFO [org.ovirt.engine.core.vdsbroker.gluster.GlusterVolumesListVDSCommand] (DefaultQuartzScheduler_Worker-19) FINISH, GlusterVolumesListVDSCommand, return: {955b86a9-10b4-463b-8555-3c321bd72f5c=org.ovirt.engine.core.common.businessentities.gluster.GlusterVolumeEntity@a57e1af5, 0bf7869b-873b-4c28-9c09-d31100f4e12a=org.ovirt.engine.core.common.businessentities.gluster.GlusterVolumeEntity@ca2565b6, 248aa48d-6aa5-4b21-867e-994265a3f145=org.ovirt.engine.core.common.businessentities.gluster.GlusterVolumeEntity@c4b83856}, log id: 32bca729 2014-06-11 14:19:16,017 INFO [org.ovirt.engine.core.vdsbroker.IsVmDuringInitiatingVDSCommand] (ajp--127.0.0.1-8702-6) [67888f4b] START, IsVmDuringInitiatingVDSCommand( vmId = ab505d9b-1811-4153-8d84-3efc7d878898), log id: 4a96a08 2014-06-11 14:19:16,018 INFO [org.ovirt.engine.core.vdsbroker.IsVmDuringInitiatingVDSCommand] (ajp--127.0.0.1-8702-6) [67888f4b] FINISH, IsVmDuringInitiatingVDSCommand, return: false, log id: 4a96a08 2014-06-11 14:19:16,088 INFO [org.ovirt.engine.core.bll.RunVmOnceCommand] (ajp--127.0.0.1-8702-6) [67888f4b] Running command: RunVmOnceCommand internal: false. Entities affected : ID: ab505d9b-1811-4153-8d84-3efc7d878898 Type: VM 2014-06-11 14:19:17,091 INFO [org.ovirt.engine.core.bll.LoginUserCommand] (ajp--127.0.0.1-8702-2) Running command: LoginUserCommand internal: false. 2014-06-11 14:19:17,096 INFO [org.ovirt.engine.core.dal.dbbroker.auditloghandling.AuditLogDirector] (ajp--127.0.0.1-8702-2) Correlation ID: null, Call Stack: null, Custom Event ID: -1, Message: User admin logged in. 2014-06-11 14:19:18,919 ERROR [org.ovirt.engine.core.bll.scheduling.external.ExternalSchedulerBrokerImpl]
Re: [ovirt-users] SLA : RAM scheduling
Hi Gilad, I followed your instructions, but it didn't work for me, I think for many reasons. Le 26/05/2014 16:22, Gilad Chaplik a écrit : Hi Nathanaël, happy to assist :) hope it will work in first run: 1) install the proxy and ovirtsdk. yum install -y ovirt-scheduler-proxy 2) put attached file in the right place (according to docs: .../plugins), according to the docs (http://www.ovirt.org/Features/oVirt_External_Scheduling_Proxy), the place should be plugins_path=$PYTHONPATH/ovirt_scheduler/plugins, which is should be mentionned in Scheduler conf file (/etc/ovirt/scheduler/scheduler.conf), but none of them exists in the filesystem!!! I've only found /usr/share/ovirt-scheduler-proxy/plugins for placing the python file and no scheduler.conf is present where I could change that path. make sure to edit the file with your ovirt's ip, user@domain and PW. My engine API can't be reached in http, so there is some work to do this with https. Here is what I did : according to /usr/lib/python2.6/site-packages/ovirtsdk/api.py, I add insecure=True to the chain connection = API(url='http*s*://host:port', username='user@domain', password=''*insecure='True'*) Maybe it is not enough and it would be useful to add validate_cert_chain=False... 3) restart proxy service. 3) use config tool to configure ovirt-engine: * ExternalSchedulerServiceURL=http://ip:18781/ The scheduler proxy listens to localhost:18781, none of ips that can be filled here will be reached on that port. * ExternalSchedulerEnabled=true 4) restart ovirt-engine service. 5) under configure-cluster_policy see that weight function memory_even_distribution was added (should be in manage policy units or /sth- you will see it in the main dialog as well). 6) clone/copy currernt cluster's used cluster policy (probably none - prefer it to have no balancing modules to avoid conflicts), name it 'your_name' and attach memory_even_distribution weight (you can leave it as the only weight module in weight section to avoid configuring factors). 7) replace cluster's cluster policy with newly created one. try it out and let me know how goes :-) Thanks, Gilad. If all above is tru, should be a good idea to update the wiki... - Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Gilad Chaplik gchap...@redhat.com Cc: Karli Sjöberg karli.sjob...@slu.se, users users@ovirt.org Sent: Monday, May 26, 2014 12:12:41 PM Subject: Re: [ovirt-users] SLA : RAM scheduling Hi Gilad, Yes, I'm interested in that feature, I'vebegun to read this following your advices http://www.ovirt.org/Features/oVirt_External_Scheduling_Proxy, I don't think I will have time enough to get it work quickly, I'm not experienced enough with this kind of feature, unless you help me. Le 24/05/2014 10:49, Gilad Chaplik a écrit : Hi Nathanaël, You have 2 ways to get what you're after (quick/slow): 1) install 'oVirt's external scheduling proxy', and write an extremely simple weight function that orders hosts by used memory, then add that to your cluster policy. 2) open an RFE for oVirt 3.4 to have that in (https://bugzilla.redhat.com/enter_bug.cgi?product=oVirt). let me know if you consider (1), and I'll assist. anyway I suggest you'll open an RFE for 3.5. Thanks, Gilad. - Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Karli Sjöberg karli.sjob...@slu.se Cc: users users@ovirt.org Sent: Friday, May 23, 2014 7:38:40 PM Subject: Re: [ovirt-users] SLA : RAM scheduling even distribution is for cpu only Le 23/05/2014 17:48, Karli Sjöberg a écrit : Den 23 maj 2014 17:13 skrev =?ISO-8859-1?Q?Nathana=EBl_Blanchet?= blanc...@abes.fr : Le 23/05/2014 17:11, Nathanaël Blanchet a écrit : Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? ... so as to both hosts be RAM evenly distributed... hope to be enough clear... Sounds like you just want to apply the cluster policy for even distribution. Have you assigned any policy for that cluster? /K -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr
Re: [ovirt-users] SLA : RAM scheduling
- Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Gilad Chaplik gchap...@redhat.com Cc: Karli Sjöberg karli.sjob...@slu.se, users users@ovirt.org Sent: Wednesday, May 28, 2014 6:26:16 PM Subject: Re: [ovirt-users] SLA : RAM scheduling Hi Gilad, I followed your instructions, but it didn't work for me, I think for many reasons. Le 26/05/2014 16:22, Gilad Chaplik a écrit : Hi Nathanaël, happy to assist :) hope it will work in first run: 1) install the proxy and ovirtsdk. yum install -y ovirt-scheduler-proxy 2) put attached file in the right place (according to docs: .../plugins), according to the docs (http://www.ovirt.org/Features/oVirt_External_Scheduling_Proxy), the place should be plugins_path=$PYTHONPATH/ovirt_scheduler/plugins, which is should be mentionned in Scheduler conf file (/etc/ovirt/scheduler/scheduler.conf), but none of them exists in the filesystem!!! I've only found /usr/share/ovirt-scheduler-proxy/plugins for placing the python file and no scheduler.conf is present where I could change that path. could be a bug in docs or installation - once we'll get it working for you I'll see to it. btw, change 'abs(0 - engine_host.max_scheduling_memory)' to 'sys.maxsize - engine_host.max_scheduling_memory' make sure to edit the file with your ovirt's ip, user@domain and PW. My engine API can't be reached in http, so there is some work to do this with https. Here is what I did : according to /usr/lib/python2.6/site-packages/ovirtsdk/api.py, I add insecure=True to the chain connection = API(url='http*s*://host:port', username='user@domain', password=''*insecure='True'*) Maybe it is not enough and it would be useful to add validate_cert_chain=False... Martin? 3) restart proxy service. 3) use config tool to configure ovirt-engine: * ExternalSchedulerServiceURL=http://ip:18781/ The scheduler proxy listens to localhost:18781, none of ips that can be filled here will be reached on that port. you config ovirt engine with the ip of the proxy, or I'm missing sth. ovirt communicates with the proxy and not the other way around. * ExternalSchedulerEnabled=true 4) restart ovirt-engine service. 5) under configure-cluster_policy see that weight function memory_even_distribution was added (should be in manage policy units or /sth- you will see it in the main dialog as well). 6) clone/copy currernt cluster's used cluster policy (probably none - prefer it to have no balancing modules to avoid conflicts), name it 'your_name' and attach memory_even_distribution weight (you can leave it as the only weight module in weight section to avoid configuring factors). 7) replace cluster's cluster policy with newly created one. try it out and let me know how goes :-) Thanks, Gilad. If all above is tru, should be a good idea to update the wiki... - Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Gilad Chaplik gchap...@redhat.com Cc: Karli Sjöberg karli.sjob...@slu.se, users users@ovirt.org Sent: Monday, May 26, 2014 12:12:41 PM Subject: Re: [ovirt-users] SLA : RAM scheduling Hi Gilad, Yes, I'm interested in that feature, I'vebegun to read this following your advices http://www.ovirt.org/Features/oVirt_External_Scheduling_Proxy, I don't think I will have time enough to get it work quickly, I'm not experienced enough with this kind of feature, unless you help me. Le 24/05/2014 10:49, Gilad Chaplik a écrit : Hi Nathanaël, You have 2 ways to get what you're after (quick/slow): 1) install 'oVirt's external scheduling proxy', and write an extremely simple weight function that orders hosts by used memory, then add that to your cluster policy. 2) open an RFE for oVirt 3.4 to have that in (https://bugzilla.redhat.com/enter_bug.cgi?product=oVirt). let me know if you consider (1), and I'll assist. anyway I suggest you'll open an RFE for 3.5. Thanks, Gilad. - Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Karli Sjöberg karli.sjob...@slu.se Cc: users users@ovirt.org Sent: Friday, May 23, 2014 7:38:40 PM Subject: Re: [ovirt-users] SLA : RAM scheduling even distribution is for cpu only Le 23/05/2014 17:48, Karli Sjöberg a écrit : Den 23 maj 2014 17:13 skrev =?ISO-8859-1?Q?Nathana=EBl_Blanchet?= blanc...@abes.fr : Le 23/05/2014 17:11, Nathanaël Blanchet a écrit : Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? ... so as to both hosts be RAM evenly distributed... hope to be enough
Re: [ovirt-users] SLA : RAM scheduling
Hi Nathanaël, happy to assist :) hope it will work in first run: 1) install the proxy and ovirtsdk. 2) put attached file in the right place (according to docs: .../plugins), make sure to edit the file with your ovirt's ip, user@domain and PW. 3) restart proxy service. 3) use config tool to configure ovirt-engine: * ExternalSchedulerServiceURL=http://ip:18781/ * ExternalSchedulerEnabled=true 4) restart ovirt-engine service. 5) under configure-cluster_policy see that weight function memory_even_distribution was added (should be in manage policy units or /sth- you will see it in the main dialog as well). 6) clone/copy currernt cluster's used cluster policy (probably none - prefer it to have no balancing modules to avoid conflicts), name it 'your_name' and attach memory_even_distribution weight (you can leave it as the only weight module in weight section to avoid configuring factors). 7) replace cluster's cluster policy with newly created one. try it out and let me know how goes :-) Thanks, Gilad. - Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Gilad Chaplik gchap...@redhat.com Cc: Karli Sjöberg karli.sjob...@slu.se, users users@ovirt.org Sent: Monday, May 26, 2014 12:12:41 PM Subject: Re: [ovirt-users] SLA : RAM scheduling Hi Gilad, Yes, I'm interested in that feature, I'vebegun to read this following your advices http://www.ovirt.org/Features/oVirt_External_Scheduling_Proxy, I don't think I will have time enough to get it work quickly, I'm not experienced enough with this kind of feature, unless you help me. Le 24/05/2014 10:49, Gilad Chaplik a écrit : Hi Nathanaël, You have 2 ways to get what you're after (quick/slow): 1) install 'oVirt's external scheduling proxy', and write an extremely simple weight function that orders hosts by used memory, then add that to your cluster policy. 2) open an RFE for oVirt 3.4 to have that in (https://bugzilla.redhat.com/enter_bug.cgi?product=oVirt). let me know if you consider (1), and I'll assist. anyway I suggest you'll open an RFE for 3.5. Thanks, Gilad. - Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Karli Sjöberg karli.sjob...@slu.se Cc: users users@ovirt.org Sent: Friday, May 23, 2014 7:38:40 PM Subject: Re: [ovirt-users] SLA : RAM scheduling even distribution is for cpu only Le 23/05/2014 17:48, Karli Sjöberg a écrit : Den 23 maj 2014 17:13 skrev =?ISO-8859-1?Q?Nathana=EBl_Blanchet?= blanc...@abes.fr : Le 23/05/2014 17:11, Nathanaël Blanchet a écrit : Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? ... so as to both hosts be RAM evenly distributed... hope to be enough clear... Sounds like you just want to apply the cluster policy for even distribution. Have you assigned any policy for that cluster? /K -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr from ovirtsdk.xml import params from ovirtsdk.api import API import sys class memory_even_distribution(): '''rank hosts by even ram distribution''' properties_validation = '' def _get_connection(self): #open a connection to the rest api connection = None try: connection = API(url='http://host:port', username='user@domain', password='') except BaseException as ex: #letting the external proxy know there was an error print sys.stderr, ex return None return connection def _get_hosts(self, host_ids, connection): #get all the hosts with the given
Re: [ovirt-users] SLA : RAM scheduling
- Original Message - From: Gilad Chaplik gchap...@redhat.com To: Nathanaël Blanchet blanc...@abes.fr Cc: users users@ovirt.org Sent: Saturday, May 24, 2014 11:52:10 AM Subject: Re: [ovirt-users] SLA : RAM scheduling - Original Message - From: Gilad Chaplik gchap...@redhat.com To: Nathanaël Blanchet blanc...@abes.fr Cc: Karli Sjöberg karli.sjob...@slu.se, users users@ovirt.org Sent: Saturday, May 24, 2014 11:49:48 AM Subject: Re: [ovirt-users] SLA : RAM scheduling Hi Nathanaël, You have 2 ways to get what you're after (quick/slow): 1) install 'oVirt's external scheduling proxy', and write an extremely simple weight function that orders hosts by used memory, then add that to your cluster policy. 2) open an RFE for oVirt 3.4 to have that in (https://bugzilla.redhat.com/enter_bug.cgi?product=oVirt). by 3.4, I mean 3.4.x (= anyway for (2) you'll need to upgrade), but not sure it will make it. let me know if you consider (1), and I'll assist. anyway I suggest you'll open an RFE for 3.5. Thanks, Gilad. - Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Karli Sjöberg karli.sjob...@slu.se Cc: users users@ovirt.org Sent: Friday, May 23, 2014 7:38:40 PM Subject: Re: [ovirt-users] SLA : RAM scheduling even distribution is for cpu only Le 23/05/2014 17:48, Karli Sjöberg a écrit : Den 23 maj 2014 17:13 skrev =?ISO-8859-1?Q?Nathana=EBl_Blanchet?= blanc...@abes.fr : Le 23/05/2014 17:11, Nathanaël Blanchet a écrit : Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? ... so as to both hosts be RAM evenly distributed... hope to be enough clear... Sounds like you just want to apply the cluster policy for even distribution. Have you assigned any policy for that cluster? /K -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr Sounds like this RFE: https://bugzilla.redhat.com/show_bug.cgi?id=1093038 FWIW, you can implement your own logic in Python until we get to implement the above RFE. ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
Hi Nathanaël, You have 2 ways to get what you're after (quick/slow): 1) install 'oVirt's external scheduling proxy', and write an extremely simple weight function that orders hosts by used memory, then add that to your cluster policy. 2) open an RFE for oVirt 3.4 to have that in (https://bugzilla.redhat.com/enter_bug.cgi?product=oVirt). let me know if you consider (1), and I'll assist. anyway I suggest you'll open an RFE for 3.5. Thanks, Gilad. - Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Karli Sjöberg karli.sjob...@slu.se Cc: users users@ovirt.org Sent: Friday, May 23, 2014 7:38:40 PM Subject: Re: [ovirt-users] SLA : RAM scheduling even distribution is for cpu only Le 23/05/2014 17:48, Karli Sjöberg a écrit : Den 23 maj 2014 17:13 skrev =?ISO-8859-1?Q?Nathana=EBl_Blanchet?= blanc...@abes.fr : Le 23/05/2014 17:11, Nathanaël Blanchet a écrit : Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? ... so as to both hosts be RAM evenly distributed... hope to be enough clear... Sounds like you just want to apply the cluster policy for even distribution. Have you assigned any policy for that cluster? /K -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
- Original Message - From: Gilad Chaplik gchap...@redhat.com To: Nathanaël Blanchet blanc...@abes.fr Cc: Karli Sjöberg karli.sjob...@slu.se, users users@ovirt.org Sent: Saturday, May 24, 2014 11:49:48 AM Subject: Re: [ovirt-users] SLA : RAM scheduling Hi Nathanaël, You have 2 ways to get what you're after (quick/slow): 1) install 'oVirt's external scheduling proxy', and write an extremely simple weight function that orders hosts by used memory, then add that to your cluster policy. 2) open an RFE for oVirt 3.4 to have that in (https://bugzilla.redhat.com/enter_bug.cgi?product=oVirt). by 3.4, I mean 3.4.x (= anyway for (2) you'll need to upgrade), but not sure it will make it. let me know if you consider (1), and I'll assist. anyway I suggest you'll open an RFE for 3.5. Thanks, Gilad. - Original Message - From: Nathanaël Blanchet blanc...@abes.fr To: Karli Sjöberg karli.sjob...@slu.se Cc: users users@ovirt.org Sent: Friday, May 23, 2014 7:38:40 PM Subject: Re: [ovirt-users] SLA : RAM scheduling even distribution is for cpu only Le 23/05/2014 17:48, Karli Sjöberg a écrit : Den 23 maj 2014 17:13 skrev =?ISO-8859-1?Q?Nathana=EBl_Blanchet?= blanc...@abes.fr : Le 23/05/2014 17:11, Nathanaël Blanchet a écrit : Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? ... so as to both hosts be RAM evenly distributed... hope to be enough clear... Sounds like you just want to apply the cluster policy for even distribution. Have you assigned any policy for that cluster? /K -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
[ovirt-users] SLA : RAM scheduling
Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
Le 23/05/2014 17:11, Nathanaël Blanchet a écrit : Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? ... so as to both hosts be RAM evenly distributed... hope to be enough clear... -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
Den 23 maj 2014 17:13 skrev =?ISO-8859-1?Q?Nathana=EBl_Blanchet?= blanc...@abes.fr: Le 23/05/2014 17:11, Nathanaël Blanchet a écrit : Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? ... so as to both hosts be RAM evenly distributed... hope to be enough clear... Sounds like you just want to apply the cluster policy for even distribution. Have you assigned any policy for that cluster? /K -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] SLA : RAM scheduling
even distribution is for cpu only Le 23/05/2014 17:48, Karli Sjöberg a écrit : Den 23 maj 2014 17:13 skrev =?ISO-8859-1?Q?Nathana=EBl_Blanchet?= blanc...@abes.fr: Le 23/05/2014 17:11, Nathanaël Blanchet a écrit : Hello, On ovirt 3.4, is it possible to schedule vms distribution depending on host RAM availibility? Concretly, I had to manually move vms all the vms to the second host of the cluster, this lead to reach 90% occupation of memory on the destination host. When my first host has rebooted, none vms of the second host automatically migrated to the first one which had full RAM. How to make this happen? ... so as to both hosts be RAM evenly distributed... hope to be enough clear... Sounds like you just want to apply the cluster policy for even distribution. Have you assigned any policy for that cluster? /K -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users -- Nathanaël Blanchet Supervision réseau Pôle exploitation et maintenance Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5 Tél. 33 (0)4 67 54 84 55 Fax 33 (0)4 67 54 84 14 blanc...@abes.fr ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users