Re: [Openstack-operators] Sharing resources across OpenStack instances
Le 22/04/2015 15:32, Adam Young a écrit : Its been my understanding that many people are deploying small OpenStack instances as a way to share the Hardware owned by their particular team, group, or department. The Keystone instance represents ownership, and the identity of the users comes from a corporate LDAP server. Is there much demand for the following scenarios? 1. A project team crosses organizational boundaries and has to work with VMs in two separate OpenStack instances. They need to set up a network that requires talking to two neutron instances. 2. One group manages a powerful storage array. Several OpenStack instances need to be able to mount volumes from this array. Sometimes, those volumes have to be transferred from VMs running in one instance to another. 3. A group is producing nightly builds. Part of this is an image building system that posts to glance. Ideally, multiple OpenStack instances would be able to pull their images from the same glance. 4. Hadoop ( or some other orchestrated task) requires more resources than are in any single OpenStack instance, and needs to allocate resources across two or more instances for a single job. I suspect that these kinds of architectures are becoming more common. Can some of the operators validate these assumptions? Are there other, more common cases where Operations need to span multiple clouds which would require integration of one Nova server with multiple Cinder, Glance, or Neutron servers managed in other OpenStack instances? I'm always a bit disappointed when someone asks me about hybridation with OpenStack. OpenStack is not a Cloud Management Platform, that can manage severeal Clouds, private and public... Like RedHat Cloudforms / ManageIQ. At least, being able to manage two OpenStack instances, one private and one from a public Cloud with OpenStack API will be great. I found a project by Huawei to cascade Openstack instances : https://wiki.openstack.org/wiki/OpenStack_cascading_solution I really would like to see that becoming reality. Perhaps It can be a solution to your scenarii ? ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] Sharing resources across OpenStack instances
This is a case for a cross project cloud (institutional?). It costs more to run two little clouds then one bigger one. Both in terms of man power, and in cases like these. under utilized resources. #3 is interesting though. If there is to be an openstack app catalog, it would be inportant to be able to pull the needed images from outside the cloud easily. Thanks, Kevin From: Adam Young Sent: Wednesday, April 22, 2015 6:32:17 AM To: openstack-operators@lists.openstack.org Subject: [Openstack-operators] Sharing resources across OpenStack instances Its been my understanding that many people are deploying small OpenStack instances as a way to share the Hardware owned by their particular team, group, or department. The Keystone instance represents ownership, and the identity of the users comes from a corporate LDAP server. Is there much demand for the following scenarios? 1. A project team crosses organizational boundaries and has to work with VMs in two separate OpenStack instances. They need to set up a network that requires talking to two neutron instances. 2. One group manages a powerful storage array. Several OpenStack instances need to be able to mount volumes from this array. Sometimes, those volumes have to be transferred from VMs running in one instance to another. 3. A group is producing nightly builds. Part of this is an image building system that posts to glance. Ideally, multiple OpenStack instances would be able to pull their images from the same glance. 4. Hadoop ( or some other orchestrated task) requires more resources than are in any single OpenStack instance, and needs to allocate resources across two or more instances for a single job. I suspect that these kinds of architectures are becoming more common. Can some of the operators validate these assumptions? Are there other, more common cases where Operations need to span multiple clouds which would require integration of one Nova server with multiple Cinder, Glance, or Neutron servers managed in other OpenStack instances? ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
[Openstack-operators] StackTach.v3 now in production ...
(sorry for cross-post, but this is appropriate to both audiences) Hey y'all! For those of you that don't know, StackTach is a notification-based debugging, monitoring and usage tool for OpenStack. We're happy to announce that we've recently rolled StackTach.v3 into production at one of the Rax datacenters with a plan to roll out to the rest asap. Once we do, we'll be bumping all the library versions to 1.0, but we encourage you to start playing with the system now. The docs and screencasts are at www.stacktach.com We live on #stacktach on freenode (for v2 and v3 questions) All the StackTach code is on stackforge https://github.com/stackforge?query=stacktach This a very exciting time for us. With StackTach.v3 we've: * solved many of the scaling, redundancy and idempotency problems of v2 * modularized the entire system (use only the parts you want) * made the system less rigid with respect to Nova and Glance. Now, nearly any JSON notification can be handled (even outside of OpenStack) * created a very flexible REST API with pluggable implementation drivers. So, if you don't like our solution but want to keep a compatible API, all the pieces are there for you, including cmdline tools and client libraries. * included a devstack-like sandbox for you to play in that doesn't require an OpenStack installation to generate notifications * developed a way to run STv3 side-by-side with your existing notification consumers for safe trials. We can split notification queues without requiring any changes to your openstack deployment (try *that* with oslo-messaging ;) If you haven't looked at your OpenStack deployment from the perspective of notifications you're really missing out. It's the most powerful way to debug your installations. And, for usage metering, there is really no better option. We feel StackTach.v3 is the best solution out there for all your event-processing needs. Let us know how we can help! We're in a good place to squash bugs quickly. Cheers -Sandy, Dragon and the rest of the StackTach.v3 team/contributors ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] over commit ratios
In addition to these factors, collocation happens to be another key source of noise. By collocation I mean VMs doing the same/similar work running on the same hypervisor. This happens under low capacity situations when the scheduler could not enforce anti-affinity. Subbu On Apr 22, 2015, at 5:09 AM, George Shuklin george.shuk...@gmail.com wrote: Yes, it really depends on the used backing technique. We using SSDs and raw images, so IO is not an issue. But memory is more important: if you lack IO capability you left with slow guests. If you lack memory you left with dead guests (hello, OOM killer). BTW: Swap is needed not to swapin/swapout, but to relief memory pressure. With properly configured memory swin/swout should be less than 2-3. On 04/22/2015 09:49 AM, Tim Bell wrote: I'd also keep an eye on local I/O... we've found this to be the resource which can cause the worst noisy neighbours. Swapping makes this worse. Tim -Original Message- From: George Shuklin [mailto:george.shuk...@gmail.com] Sent: 21 April 2015 23:55 To: openstack-operators@lists.openstack.org Subject: Re: [Openstack-operators] over commit ratios It's very depend on production type. If you can control guests and predict their memory consumption, use it as base for ratio. If you can't (typical for public clouds) - use 1 or smaller with reserved_host_memory_mb in nova.conf. And one more: some swap sapce is really necessary. Add at least twice of reserved_host_memory_mb - it really improves performance and prevents strange OOMs in the situation of very large host with very small dom0 footprint. On 04/21/2015 10:59 PM, Caius Howcroft wrote: Just a general question: what kind of over commit ratios do people normally run in production with? We currently run 2 for cpu and 1 for memory (with some held back for OS/ceph) i.e.: default['bcpc']['nova']['ram_allocation_ratio'] = 1.0 default['bcpc']['nova']['reserved_host_memory_mb'] = 1024 # often larger default['bcpc']['nova']['cpu_allocation_ratio'] = 2.0 Caius ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] Operator Help to get patch merged
I just read this email and after looking what happened on gerrit, I feel totally disappointed of the way the Horizon team (specially the cores) handle this patch. You did everything properly in your commit and horizon dev team should have helped you to get your commit merge. I do not understand why you are not even included as co-author in the patch that was finally merged: https://review.openstack.org/#/c/175702/ Just my two cents! Edgar From: Joseph Bajin josephba...@gmail.commailto:josephba...@gmail.com Date: Saturday, April 18, 2015 at 5:06 PM To: OpenStack Operators openstack-operators@lists.openstack.orgmailto:openstack-operators@lists.openstack.org Subject: [Openstack-operators] Operator Help to get patch merged I wanted to see about getting some Operator Help to push through a patch[1]. The patch is to not give the user a 404 message back when they click the cancel button why trying to change their password or the user settings. The patch resets the page. It's been sitting there for a while, but started to get some -1's and then +1's, then some move to kilo-rc1 to liberty and back. Some people think that those screens should be somewhere else, others think the text should be replaced, but that is not the purpose of the patch. It's just to not give back a negative user experience. So, I'm hoping that I can get some Operator support to get this merged and if they want to change the text, change the location, etc, then they can do it later down the road. Thanks -Joe [1] https://review.openstack.org/#/c/166569/ ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] over commit ratios
A juno feature may help with this, Utilization based scheduling: https://blueprints.launchpad.net/nova/+spec/utilization-aware-scheduling That helps when landing the instance, but doesn't help if utilization changes /after/ instances have landed, but could help with a resize action to relocate the instance. - jlk On Wed, Apr 22, 2015 at 8:44 AM, Allamaraju, Subbu su...@subbu.org wrote: In addition to these factors, collocation happens to be another key source of noise. By collocation I mean VMs doing the same/similar work running on the same hypervisor. This happens under low capacity situations when the scheduler could not enforce anti-affinity. Subbu On Apr 22, 2015, at 5:09 AM, George Shuklin george.shuk...@gmail.com wrote: Yes, it really depends on the used backing technique. We using SSDs and raw images, so IO is not an issue. But memory is more important: if you lack IO capability you left with slow guests. If you lack memory you left with dead guests (hello, OOM killer). BTW: Swap is needed not to swapin/swapout, but to relief memory pressure. With properly configured memory swin/swout should be less than 2-3. On 04/22/2015 09:49 AM, Tim Bell wrote: I'd also keep an eye on local I/O... we've found this to be the resource which can cause the worst noisy neighbours. Swapping makes this worse. Tim -Original Message- From: George Shuklin [mailto:george.shuk...@gmail.com] Sent: 21 April 2015 23:55 To: openstack-operators@lists.openstack.org Subject: Re: [Openstack-operators] over commit ratios It's very depend on production type. If you can control guests and predict their memory consumption, use it as base for ratio. If you can't (typical for public clouds) - use 1 or smaller with reserved_host_memory_mb in nova.conf. And one more: some swap sapce is really necessary. Add at least twice of reserved_host_memory_mb - it really improves performance and prevents strange OOMs in the situation of very large host with very small dom0 footprint. On 04/21/2015 10:59 PM, Caius Howcroft wrote: Just a general question: what kind of over commit ratios do people normally run in production with? We currently run 2 for cpu and 1 for memory (with some held back for OS/ceph) i.e.: default['bcpc']['nova']['ram_allocation_ratio'] = 1.0 default['bcpc']['nova']['reserved_host_memory_mb'] = 1024 # often larger default['bcpc']['nova']['cpu_allocation_ratio'] = 2.0 Caius ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
[Openstack-operators] [all][api] 3 API Guidelines up for final review
Hi All, We have 3 API Guidelines that are ready for a final review. 1. Metadata guidelines document https://review.openstack.org/#/c/141229/ 2. Tagging guidelines https://review.openstack.org/#/c/155620/ 3. Guidelines on using date and time format https://review.openstack.org/#/c/159892/ If the API Working Group hasn’t received any further feedback, we’ll merge them on April 29. Thanks, Everett ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
[Openstack-operators] [Ceilometer] Gnocchi - capturing time-series data
hi folks, there's a session coming up at the summit so we can discuss Ceilometer and give/get some feedback but i wanted to highlight some of the work we've been doing, specifically relating to storing measurement values. as many of you have heard/read, we're building this thing called Gnocchi. to explain what Gnocchi is, Julien, a key developer of Gnocchi, wrote up an introduction[1] which i hope explains a bit of the concepts of Gnocchi and what it aims to achieve. from this, i hope it answers some questions, generates some ideas on how it can be used, and raises a few thoughts on how to improve it. [1] http://julien.danjou.info/blog/2015/openstack-gnocchi-first-release cheers, gord ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] Sharing resources across OpenStack instances
[top posting on this one] Hi, When you write Openstack instances, I'm assuming that you're referring to Openstack deployments right? We have different deployments based on geographic regions for performance concerns but certainly not by department. Each Openstack project is tied to a department/project budget code and re-billed accordingly based on Ceilometer data. No need to have separate deployments for that. Central IT owns all the Cloud infra. In the separate deployments the only thing that we aim to have shared is Swift and Keystone (it's not the case for us right now). Glance images need to be identical between deployments but that's easily achievable through automation both for the operator and the end user. We make sure that the users understand that these are separate regions/Clouds akin to AWS regions. -m On Wed, 2015-04-22 at 13:50 +, Fox, Kevin M wrote: This is a case for a cross project cloud (institutional?). It costs more to run two little clouds then one bigger one. Both in terms of man power, and in cases like these. under utilized resources. #3 is interesting though. If there is to be an openstack app catalog, it would be inportant to be able to pull the needed images from outside the cloud easily. Thanks, Kevin __ From: Adam Young Sent: Wednesday, April 22, 2015 6:32:17 AM To: openstack-operators@lists.openstack.org Subject: [Openstack-operators] Sharing resources across OpenStack instances Its been my understanding that many people are deploying small OpenStack instances as a way to share the Hardware owned by their particular team, group, or department. The Keystone instance represents ownership, and the identity of the users comes from a corporate LDAP server. Is there much demand for the following scenarios? 1. A project team crosses organizational boundaries and has to work with VMs in two separate OpenStack instances. They need to set up a network that requires talking to two neutron instances. 2. One group manages a powerful storage array. Several OpenStack instances need to be able to mount volumes from this array. Sometimes, those volumes have to be transferred from VMs running in one instance to another. 3. A group is producing nightly builds. Part of this is an image building system that posts to glance. Ideally, multiple OpenStack instances would be able to pull their images from the same glance. 4. Hadoop ( or some other orchestrated task) requires more resources than are in any single OpenStack instance, and needs to allocate resources across two or more instances for a single job. I suspect that these kinds of architectures are becoming more common. Can some of the operators validate these assumptions? Are there other, more common cases where Operations need to span multiple clouds which would require integration of one Nova server with multiple Cinder, Glance, or Neutron servers managed in other OpenStack instances? ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] Sharing resources across OpenStack instances
At Overstock we have a number of separate Openstack deployments in different facilities that are completely separated from each other. No shared services between them. Some of the separation is due to the kind of instances they contain (“Dev/Test vs “Prod” for example), but it is largely due to the location diversity and the desire to not require everything to be upgraded at once. We have automation to build and push out new glance images every month (built via Oz) to the multiple glance instances. Our home-made orchestration knows how to provision to the multiple clouds. We are not currently using corporate LDAP for these, but our orchestration tool does use AD and that’s where we do most of our work from anyway. All of these are managed by our Website Systems team and we do basic show-back to various teams that use these services using data from nova statistics (no Ceilometer yet). Mike Smith Principal Engineer, Website Systems Overstock.com On Apr 22, 2015, at 6:26 PM, Marc Heckmann marc.heckm...@ubisoft.com wrote: [top posting on this one] Hi, When you write Openstack instances, I'm assuming that you're referring to Openstack deployments right? We have different deployments based on geographic regions for performance concerns but certainly not by department. Each Openstack project is tied to a department/project budget code and re-billed accordingly based on Ceilometer data. No need to have separate deployments for that. Central IT owns all the Cloud infra. In the separate deployments the only thing that we aim to have shared is Swift and Keystone (it's not the case for us right now). Glance images need to be identical between deployments but that's easily achievable through automation both for the operator and the end user. We make sure that the users understand that these are separate regions/Clouds akin to AWS regions. -m On Wed, 2015-04-22 at 13:50 +, Fox, Kevin M wrote: This is a case for a cross project cloud (institutional?). It costs more to run two little clouds then one bigger one. Both in terms of man power, and in cases like these. under utilized resources. #3 is interesting though. If there is to be an openstack app catalog, it would be inportant to be able to pull the needed images from outside the cloud easily. Thanks, Kevin __ From: Adam Young Sent: Wednesday, April 22, 2015 6:32:17 AM To: openstack-operators@lists.openstack.org Subject: [Openstack-operators] Sharing resources across OpenStack instances Its been my understanding that many people are deploying small OpenStack instances as a way to share the Hardware owned by their particular team, group, or department. The Keystone instance represents ownership, and the identity of the users comes from a corporate LDAP server. Is there much demand for the following scenarios? 1. A project team crosses organizational boundaries and has to work with VMs in two separate OpenStack instances. They need to set up a network that requires talking to two neutron instances. 2. One group manages a powerful storage array. Several OpenStack instances need to be able to mount volumes from this array. Sometimes, those volumes have to be transferred from VMs running in one instance to another. 3. A group is producing nightly builds. Part of this is an image building system that posts to glance. Ideally, multiple OpenStack instances would be able to pull their images from the same glance. 4. Hadoop ( or some other orchestrated task) requires more resources than are in any single OpenStack instance, and needs to allocate resources across two or more instances for a single job. I suspect that these kinds of architectures are becoming more common. Can some of the operators validate these assumptions? Are there other, more common cases where Operations need to span multiple clouds which would require integration of one Nova server with multiple Cinder, Glance, or Neutron servers managed in other OpenStack instances? ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators CONFIDENTIALITY NOTICE: This message is intended only for the use and review of the individual or entity to which it is addressed and may contain information that is privileged and confidential. If the reader of this message is not the
Re: [Openstack-operators] over commit ratios
I'd also keep an eye on local I/O... we've found this to be the resource which can cause the worst noisy neighbours. Swapping makes this worse. Tim -Original Message- From: George Shuklin [mailto:george.shuk...@gmail.com] Sent: 21 April 2015 23:55 To: openstack-operators@lists.openstack.org Subject: Re: [Openstack-operators] over commit ratios It's very depend on production type. If you can control guests and predict their memory consumption, use it as base for ratio. If you can't (typical for public clouds) - use 1 or smaller with reserved_host_memory_mb in nova.conf. And one more: some swap sapce is really necessary. Add at least twice of reserved_host_memory_mb - it really improves performance and prevents strange OOMs in the situation of very large host with very small dom0 footprint. On 04/21/2015 10:59 PM, Caius Howcroft wrote: Just a general question: what kind of over commit ratios do people normally run in production with? We currently run 2 for cpu and 1 for memory (with some held back for OS/ceph) i.e.: default['bcpc']['nova']['ram_allocation_ratio'] = 1.0 default['bcpc']['nova']['reserved_host_memory_mb'] = 1024 # often larger default['bcpc']['nova']['cpu_allocation_ratio'] = 2.0 Caius ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
[Openstack-operators] Migrating to Different Availability Zone and Different Storage
Hello operators, We have a running installation of Openstack with block storage. We are currently having a problem with our storage appliance and we would like to migrate the instance from this storage appliance. To do that, we are thinking of creating a new availability zone with new storage applience. In the end, we want to move the instances from our current storage appliance to another. Those storage appliances are the same and we have enough storage. Basically, the storage appliance and the interface will be the same, but it will be completely new machine. Is it possible to move instances from one availability zone to other while keeping in mind that those availability zones use different storage? If so, is there any guide to configure this? Does nova/cinder copy the disk of the instance from old one (old availability zone) to new one? Lastly, could it be done with minimal/zero downtime (live migration)? Regards, Eren -- Eren Türkay, System Administrator https://skyatlas.com/ | +90 850 885 0357 Yildiz Teknik Universitesi Davutpasa Kampusu Teknopark Bolgesi, D2 Blok No:107 Esenler, Istanbul Pk.34220 signature.asc Description: OpenPGP digital signature ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] [openstack-dev] [Nova] Add config option for real deletes instead of soft-deletes
Cross-posting to operators@ as I think they are rather interested in the $subject :-) Le 21/04/2015 23:42, Artom Lifshitz a écrit : Hello, I'd like to gauge acceptance of introducing a feature that would give operators a config option to perform real database deletes instead of soft deletes. There's definitely a need for *something* that cleans up the database. There have been a few attempts at a DB purge engine [1][2][3][4][5], and archiving to shadow tables has been merged [6] (though that currently has some issues [7]). DB archiving notwithstanding, the general response to operators when they mention the database becoming too big seems to be DIY cleanup. I would like to propose a different approach: add a config option that turns soft-deletes into real deletes, and start telling operators if you turn this on, it's DIY backups. Would something like that be acceptable and feasible? I'm ready to put in the work to implement this, however searching the mailing list indicates that it would be somewhere between non trivial and impossible [8]. Before I start, I would like some confidence that it's closer to the former than the latter :) My personal bet is that it should indeed be config-driven : either we keep the old records or we just get rid of them. I'm not fan of any massive deletion system that Nova would manage unless it would be on trigger because even if the model is quite correctly normalized, it would perhaps lock many tables if we don't get attention to that. Anyway, a spec seems a good place for discussing that, IMHO. -Sylvain Cheers! [1] https://blueprints.launchpad.net/nova/+spec/db-purge-engine [2] https://blueprints.launchpad.net/nova/+spec/db-purge2 [3] https://blueprints.launchpad.net/nova/+spec/remove-db-archiving [4] https://blueprints.launchpad.net/nova/+spec/database-purge [5] https://blueprints.launchpad.net/nova/+spec/db-archiving [6] https://review.openstack.org/#/c/18493/ [7] https://review.openstack.org/#/c/109201/ [8] http://lists.openstack.org/pipermail/openstack-operators/2014-November/005591.html __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] over commit ratios
Yes, it really depends on the used backing technique. We using SSDs and raw images, so IO is not an issue. But memory is more important: if you lack IO capability you left with slow guests. If you lack memory you left with dead guests (hello, OOM killer). BTW: Swap is needed not to swapin/swapout, but to relief memory pressure. With properly configured memory swin/swout should be less than 2-3. On 04/22/2015 09:49 AM, Tim Bell wrote: I'd also keep an eye on local I/O... we've found this to be the resource which can cause the worst noisy neighbours. Swapping makes this worse. Tim -Original Message- From: George Shuklin [mailto:george.shuk...@gmail.com] Sent: 21 April 2015 23:55 To: openstack-operators@lists.openstack.org Subject: Re: [Openstack-operators] over commit ratios It's very depend on production type. If you can control guests and predict their memory consumption, use it as base for ratio. If you can't (typical for public clouds) - use 1 or smaller with reserved_host_memory_mb in nova.conf. And one more: some swap sapce is really necessary. Add at least twice of reserved_host_memory_mb - it really improves performance and prevents strange OOMs in the situation of very large host with very small dom0 footprint. On 04/21/2015 10:59 PM, Caius Howcroft wrote: Just a general question: what kind of over commit ratios do people normally run in production with? We currently run 2 for cpu and 1 for memory (with some held back for OS/ceph) i.e.: default['bcpc']['nova']['ram_allocation_ratio'] = 1.0 default['bcpc']['nova']['reserved_host_memory_mb'] = 1024 # often larger default['bcpc']['nova']['cpu_allocation_ratio'] = 2.0 Caius ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators