What’s New in Mirantis OpenStack 8.0 Webinar Q&A, Part 2
Nick Chase - April 15, 2016
As promised, here's the second part of our Q&A from the Mirantis OpenStack 8.0 webinar. Enjoy!
Q: How do you get the Nagios options?
A: Alerting with Nagios is possible through StackLight.
Q: How do we get our plugins validated?
A: Fill out the form in Partner section of mirantis.com and we will be in touch with you within 24 hours.
Q: With Fuel Plug-ins and Day 2 to Day N support being critical, will Fuel server now support HA? As i understand it - today the Fuel server does not support HA configuration?
A: The Fuel Master node itself is not deployed in an HA fashion, however it can deploy OpenStack in an HA architecture (i.e., 3x Controllers, etc). The Fuel Master node is not in the data path, nor does it host any services required by the OpenStack deployments. Therefore, if the Fuel Master node fails, OpenStack services are unaffected. Standard backup and recovery procedures are sufficient for the Fuel Master node.
Q: In the Mirantis distro, the hosts are running Ubuntu hypervisor. We are using RHEL on our guest VMs. We are told that RHEL will not give support for guest VMs if the host hypervisor is Ubuntu. Red hat is saying that we need to reproduce the problem on a RHEL hypervisor before they look at the snag report. Is there a plan from Mirantis to improve this?
A: While you can't bootstrap RHEL KVM hosts with Fuel, you can have MOS 8.0 controllers orchestrate RHEL KVM nodes provisioned manually or via Red Hat Satellite. Red Hat not supporting RHEL VMs on top of Ubuntu KVM hosts is a RHEL support policy issue unrelated to Mirantis or our product. Mirantis ships OpenStack controllers that we test and certify to interoperate with a variety of host hypervisors, including Ubuntu KVM, RHEL KVM and vCenter/ESXi. If you want to run RHEL VMs that are officially supported by Red Hat, we recommend you make sure that the host hypervisor you use with MOS is the kind that is supported by Red Hat.
In addition, since this is a matter of Guest OS & Hypervisor interop/support issue, our understanding is that Red Hat's policy is to provide best-effort support for this combination https://access.redhat.com/articles/1067.
Q: What is the Host OS for the Mirantis Openstack administration servers? Is it Mirantis' own linux distro?
A: The Mirantis controller nodes are natively built on Ubuntu.
Q: What was the previous webinar that talked about the various components (Nova, Cinder, Neutron, etc.) that was mentioned at the first of the presentation? I'm new to OpenStack and would like to watch videos that help me understand the components.
A: We were talking about What’s New in OpenStack Liberty.
Q: Do we still have to PXE boot the nodes and deploy the OpenStack packages via Fuel or is there a plan to have a push-based OpenStack installation? Let's say I have a RHEL7 Host, how I integrate it with Fuel, say with RHEL7 host as a Compute Host ?
A: PXE boot // discover & provision // deploy is still the normal workflow in Fuel, but we have supported separate provision and deploy for several releases now. In adding support for RHEL Compute hosts, this process has been refined some. In this case you provide the provisioned RHEL node and we add Mirantis Openstack to it.
Q: How do I add more new nodes into a running OpenStack cloud in production? How many ways can do that? (I mean add more bare metal machines into an existing OpenStack deployment.)
A: If the environment was already deployed with Fuel, we have supported adding additional nodes in various roles since the very beginning. You can do this via the UI, the CLI and the Fuel API.
Q: Are you using an orchestrator (Heat) for scale out and workload management ?
A: Given the context of this question it seems to be in regards to the autoscaling of Murano Environments within Kubernetes clusters; in this case, the answer is no. The autoscaling monitors and triggers that orchestrate the scaling up and down of Kubernetes minion nodes are based on new intelligence built into the plugins that leverage Kube monitors within the Kubernetes cluster master nodes.
Q: How can I protect my OpenStack private cloud running Kubernetes from a DDoS attack that will extend VMs to the public cloud and blast the limit of my credit card?
A: You can set limits for both private cloud VMs and public cloud VMs; your app may stall, but at least you won't be hit with any unpleasant financial surprises.
Q: Are the Ironic servers counted in the Mirantis support agreement?
A: Ironic nodes are counted in the same manner as other OpenStack nodes.
Q: Can I access Horizon from an external network?
A: Yes, given the appropriate access and network routing.
Q: What features get enabled if Advanced and Mirantis features are enabled during the installation of Fuel in the setup menu?
A: The reduced footprint feature introduced in 7.0 becomes visible with Advanced. The documentation is located here: https://docs.mirantis.com/openstack/fuel/fuel-8.0/operations.html#using-the-reduced-footprint-feature
Q: Can you elaborate more on the Kubernetes feature? I mean you are not talinkng about Magnum alone or fused with Google Container Engine, but only an installation of Kubernetes via murano running on OpenStack (like any app) integrated seamlessly with GCE.
A: See this video for more information and a demo.
Q: Is Ironic compatible with Cinder?
A: No. The Liberty release of Ironic does not support Cinder volumes.
Q: Can I have a detailed installation howto for Fuel and Ironic for the Mirantis installation bare metal.
A: See https://docs.mirantis.com/openstack/fuel/fuel-8.0/user-guide.html#introduction-to-the-user-guide
Q: Ironic : do you equally support node cleaning, BMC reset to default and firmware reset for all the major hardware vendors?
A: This is dependent upon the Ironic driver for each specific vendor.
Q: How do I expand the public/floating IP range after deployment? Do you still require public/floating IP in the same CIDR?
A: Floating IPs need to be part of the Public CIDR. The floating IP range can be expanded after deployment.
Q: How do we brand the UI?
A: Currently, there is no pluggable way to brand the UI. Please contact us.
Q: Customizing & adding UI features?
A: It is possible to customize the UI by developing a Fuel plugin. Find more information about how to develop, test and publish your plugin here: https://wiki.openstack.org/wiki/Fuel/Plugins
Q: Is the GCE bursting something Mirantis implemented, or is that built into Kubernetes?
A: It was actually developed by a partner company, Biarca, by utilizing Murano, Kubernetes Cluster, and OpenVPN. The functionality is in a new Murano package that deploys Kubernetes, not something that is built into Kubernetes.
Q: Any plans for erasure coding support in ceph?
A: We continue to monitor advanced features of projects, but do not have immediate plans to support erasure coding in Ceph with MOS 8.0. We are working with the Ceph community upstream for hardware acceleration of erasure coding. While it's not currently available by default in the product, our services organization can assist getting it deployed.
Q: As you people mentioned, 170+ plugins are available with MOS. How can a community developer can register/validate developed fuel plugin for a openstack service with MOS?
A: Please see the Mirantis Unlocked partner program.
Q: What tools are recommended to create base images for OpenStack?
A: Please see the Openstack.org Image Guide.
Q: For Day 2 operations, has the Nova evacuation (live/block migration) been improved for stability? Having the ability to conduct maintenance on the compute nodes has been a tough task for us to date.
A: Bug fixes are continuing for the Mitaka release, planned for MOS 9.0.
Q: Is DPDK supported in this release? Is there anything new in this release?
A: DPDK is currently not supported in MOS 8.0.
Q: When will the "upgrade package" for 8.0 be available?
A: We are currently finalizing the upgrade procedure for MOS 8.0. Documents will be published here: https://docs.mirantis.com/openstack/fuel/fuel-8.0/release-notes.html#how-to-upgrade-the-product when they are complete in the near future.
Q: Do you support importing an existing Ceph cluster?
A: This is currently not supported in MOS 8.0.
Q: Is there a link describing how to non-destructively upgrade from previous Mirantis releases?
A: Yes, you can see a video here, and documentation is coming.
Q: It might have been covered already since I logged in a tad late, but what is activated when I select Experimental features and Advanced features when installing FUEL in MOS8?
A: The following features are experimental:
A: We are continuing work upstream with the Ironic community and we will implement these features as they become available.
Q: I can't find a LBaaS Fuel Plugin for 7.0/8.0, Will this be available anytime soon?
A: LBaaS Fuel Plugin was introduced in 6.1 as a technology preview. Currently, the recommendation for LBaaS is to use 3rd party implementations that are integrated with OpenStack Neutron advanced services.
Q: What is the difference between the functions of Ironic in 8.0 and old versions.
A: Ironic was not available in prior versions of Mirantis OpenStack.
Q: Will you cover integration of Mirantis Openstack with SDN solutions in a future webinar?
A: That sounds like a great idea! We'll look into putting something together on that; watch your email. (Got another idea? Add it in the comments!)
Q: There are many new OpenStack components that are coming up every year. How are you planning to enhance Fuel to support them?
A: We tend to pick up core features as they become stable, this typically means that we do not implement support for one or more cycles after their introduction into OpenStack. Users interested in enabling these features ahead of MOS may want to examine Fuel's plugin Framework to enable these before their formal support. (Keep in mind that this is a general statement and not a promise of supporting any particular feature. Our product management department made us add this. -- Ed.)
Q: Can we deploy Fuel 8 and OpenStack in a lab without IPMI interfaces?
A: Fuel does not rely on IPMI control of the nodes, after you configure PXE and boot the nodes against the Fuel node, you will typically not need to interact with the node's console.
Q: Is it possible to get any access to a demo, where i could get to know new versions better?
A: You can join our meet the expert sessions, or simply download Mirantis OpenStack 8.0.
Q: I'm interested in the OpenContrail plugin, but there are no updates for it since MOS6.1.
A: The latest Contrail plugin support is for the Kilo release, MOS 7.0. Mirantis is not directly working on an OpenContrail plugin. Community developers may be working on something, however.
Q: Can't Kubernetes be used for auto scaling of instances also?
A: Yes, the autoscaling functions of Kubernetes handles launches and removes instances within the OpenStack environment to support the required Kube minions supporting the workloads. This includes in the public GCE environment.
Q: Does Murano support Baremetal nodes for additional SW provisioning?
A: No. Murano does not support Ironic nodes out-of-the-box.
Q: How do you push to your subscribers the patches to your distribution? Are they notified in some automatic way?
A: Yes, registered users receive email notifications of technical bulletins. Also, they are posted on our support site: https://software.mirantis.com/mirantis-openstack-technical-bulletins/.
Q: Can you explain more about the Ceph H release used vs. the one used in Fuel 7? What about performance and erasure coding?
A: MOS 7 uses Ceph Firefly, while MOS 8 uses Ceph Hammer. We curate the individual features for hardening and stability. Please refer to the release notes for support features.
Q: Why are the the upgrade paths documented as "Experimental"? Can you guys break out the Upgrade path and documentation in a way that is easier to locate vs having it tucked into a Admin guide?
A: Upgrades have many aspects that need to be weighed to appropriately guide the processes, including environment customizations, number and types of plugins, features used, and your organizations SLAs. We provide a Professional Services offering to help you take these aspects into account and identify the best path to take.
For more information or to try it out yourself, please download Mirantis OpenStack 8.0.
Q: How do you get the Nagios options?
A: Alerting with Nagios is possible through StackLight.
Q: How do we get our plugins validated?
A: Fill out the form in Partner section of mirantis.com and we will be in touch with you within 24 hours.
Q: With Fuel Plug-ins and Day 2 to Day N support being critical, will Fuel server now support HA? As i understand it - today the Fuel server does not support HA configuration?
A: The Fuel Master node itself is not deployed in an HA fashion, however it can deploy OpenStack in an HA architecture (i.e., 3x Controllers, etc). The Fuel Master node is not in the data path, nor does it host any services required by the OpenStack deployments. Therefore, if the Fuel Master node fails, OpenStack services are unaffected. Standard backup and recovery procedures are sufficient for the Fuel Master node.
Q: In the Mirantis distro, the hosts are running Ubuntu hypervisor. We are using RHEL on our guest VMs. We are told that RHEL will not give support for guest VMs if the host hypervisor is Ubuntu. Red hat is saying that we need to reproduce the problem on a RHEL hypervisor before they look at the snag report. Is there a plan from Mirantis to improve this?
A: While you can't bootstrap RHEL KVM hosts with Fuel, you can have MOS 8.0 controllers orchestrate RHEL KVM nodes provisioned manually or via Red Hat Satellite. Red Hat not supporting RHEL VMs on top of Ubuntu KVM hosts is a RHEL support policy issue unrelated to Mirantis or our product. Mirantis ships OpenStack controllers that we test and certify to interoperate with a variety of host hypervisors, including Ubuntu KVM, RHEL KVM and vCenter/ESXi. If you want to run RHEL VMs that are officially supported by Red Hat, we recommend you make sure that the host hypervisor you use with MOS is the kind that is supported by Red Hat.
In addition, since this is a matter of Guest OS & Hypervisor interop/support issue, our understanding is that Red Hat's policy is to provide best-effort support for this combination https://access.redhat.com/articles/1067.
Q: What is the Host OS for the Mirantis Openstack administration servers? Is it Mirantis' own linux distro?
A: The Mirantis controller nodes are natively built on Ubuntu.
Q: What was the previous webinar that talked about the various components (Nova, Cinder, Neutron, etc.) that was mentioned at the first of the presentation? I'm new to OpenStack and would like to watch videos that help me understand the components.
A: We were talking about What’s New in OpenStack Liberty.
Q: Do we still have to PXE boot the nodes and deploy the OpenStack packages via Fuel or is there a plan to have a push-based OpenStack installation? Let's say I have a RHEL7 Host, how I integrate it with Fuel, say with RHEL7 host as a Compute Host ?
A: PXE boot // discover & provision // deploy is still the normal workflow in Fuel, but we have supported separate provision and deploy for several releases now. In adding support for RHEL Compute hosts, this process has been refined some. In this case you provide the provisioned RHEL node and we add Mirantis Openstack to it.
Q: How do I add more new nodes into a running OpenStack cloud in production? How many ways can do that? (I mean add more bare metal machines into an existing OpenStack deployment.)
A: If the environment was already deployed with Fuel, we have supported adding additional nodes in various roles since the very beginning. You can do this via the UI, the CLI and the Fuel API.
Q: Are you using an orchestrator (Heat) for scale out and workload management ?
A: Given the context of this question it seems to be in regards to the autoscaling of Murano Environments within Kubernetes clusters; in this case, the answer is no. The autoscaling monitors and triggers that orchestrate the scaling up and down of Kubernetes minion nodes are based on new intelligence built into the plugins that leverage Kube monitors within the Kubernetes cluster master nodes.
Q: How can I protect my OpenStack private cloud running Kubernetes from a DDoS attack that will extend VMs to the public cloud and blast the limit of my credit card?
A: You can set limits for both private cloud VMs and public cloud VMs; your app may stall, but at least you won't be hit with any unpleasant financial surprises.
Q: Are the Ironic servers counted in the Mirantis support agreement?
A: Ironic nodes are counted in the same manner as other OpenStack nodes.
Q: Can I access Horizon from an external network?
A: Yes, given the appropriate access and network routing.
Q: What features get enabled if Advanced and Mirantis features are enabled during the installation of Fuel in the setup menu?
A: The reduced footprint feature introduced in 7.0 becomes visible with Advanced. The documentation is located here: https://docs.mirantis.com/openstack/fuel/fuel-8.0/operations.html#using-the-reduced-footprint-feature
Q: Can you elaborate more on the Kubernetes feature? I mean you are not talinkng about Magnum alone or fused with Google Container Engine, but only an installation of Kubernetes via murano running on OpenStack (like any app) integrated seamlessly with GCE.
A: See this video for more information and a demo.
Q: Is Ironic compatible with Cinder?
A: No. The Liberty release of Ironic does not support Cinder volumes.
Q: Can I have a detailed installation howto for Fuel and Ironic for the Mirantis installation bare metal.
A: See https://docs.mirantis.com/openstack/fuel/fuel-8.0/user-guide.html#introduction-to-the-user-guide
Q: Ironic : do you equally support node cleaning, BMC reset to default and firmware reset for all the major hardware vendors?
A: This is dependent upon the Ironic driver for each specific vendor.
Q: How do I expand the public/floating IP range after deployment? Do you still require public/floating IP in the same CIDR?
A: Floating IPs need to be part of the Public CIDR. The floating IP range can be expanded after deployment.
Q: How do we brand the UI?
A: Currently, there is no pluggable way to brand the UI. Please contact us.
Q: Customizing & adding UI features?
A: It is possible to customize the UI by developing a Fuel plugin. Find more information about how to develop, test and publish your plugin here: https://wiki.openstack.org/wiki/Fuel/Plugins
Q: Is the GCE bursting something Mirantis implemented, or is that built into Kubernetes?
A: It was actually developed by a partner company, Biarca, by utilizing Murano, Kubernetes Cluster, and OpenVPN. The functionality is in a new Murano package that deploys Kubernetes, not something that is built into Kubernetes.
Q: Any plans for erasure coding support in ceph?
A: We continue to monitor advanced features of projects, but do not have immediate plans to support erasure coding in Ceph with MOS 8.0. We are working with the Ceph community upstream for hardware acceleration of erasure coding. While it's not currently available by default in the product, our services organization can assist getting it deployed.
Q: As you people mentioned, 170+ plugins are available with MOS. How can a community developer can register/validate developed fuel plugin for a openstack service with MOS?
A: Please see the Mirantis Unlocked partner program.
Q: What tools are recommended to create base images for OpenStack?
A: Please see the Openstack.org Image Guide.
Q: For Day 2 operations, has the Nova evacuation (live/block migration) been improved for stability? Having the ability to conduct maintenance on the compute nodes has been a tough task for us to date.
A: Bug fixes are continuing for the Mitaka release, planned for MOS 9.0.
Q: Is DPDK supported in this release? Is there anything new in this release?
A: DPDK is currently not supported in MOS 8.0.
Q: When will the "upgrade package" for 8.0 be available?
A: We are currently finalizing the upgrade procedure for MOS 8.0. Documents will be published here: https://docs.mirantis.com/openstack/fuel/fuel-8.0/release-notes.html#how-to-upgrade-the-product when they are complete in the near future.
Q: Do you support importing an existing Ceph cluster?
A: This is currently not supported in MOS 8.0.
Q: Is there a link describing how to non-destructively upgrade from previous Mirantis releases?
A: Yes, you can see a video here, and documentation is coming.
Q: It might have been covered already since I logged in a tad late, but what is activated when I select Experimental features and Advanced features when installing FUEL in MOS8?
A: The following features are experimental:
- Bonding types: "balance-xor", "broadcast", "balance-tlb", "balance-alb"
- murano-cfapi: "Install Murano service broker for Cloud Foundry"
- murano_glance_artifacts_plugin: "Enable glance artifact repository"
- Install the Fedora 3.10 longterm kernel instead of the default 2.6.32 kernel. This kernel may not work with all hardware platforms, use with caution.
- Reduced footprint: Make available possibilities to spawn vms on this [Fuel Master] node that can be assign as a normal nodes.
A: We are continuing work upstream with the Ironic community and we will implement these features as they become available.
Q: I can't find a LBaaS Fuel Plugin for 7.0/8.0, Will this be available anytime soon?
A: LBaaS Fuel Plugin was introduced in 6.1 as a technology preview. Currently, the recommendation for LBaaS is to use 3rd party implementations that are integrated with OpenStack Neutron advanced services.
Q: What is the difference between the functions of Ironic in 8.0 and old versions.
A: Ironic was not available in prior versions of Mirantis OpenStack.
Q: Will you cover integration of Mirantis Openstack with SDN solutions in a future webinar?
A: That sounds like a great idea! We'll look into putting something together on that; watch your email. (Got another idea? Add it in the comments!)
Q: There are many new OpenStack components that are coming up every year. How are you planning to enhance Fuel to support them?
A: We tend to pick up core features as they become stable, this typically means that we do not implement support for one or more cycles after their introduction into OpenStack. Users interested in enabling these features ahead of MOS may want to examine Fuel's plugin Framework to enable these before their formal support. (Keep in mind that this is a general statement and not a promise of supporting any particular feature. Our product management department made us add this. -- Ed.)
Q: Can we deploy Fuel 8 and OpenStack in a lab without IPMI interfaces?
A: Fuel does not rely on IPMI control of the nodes, after you configure PXE and boot the nodes against the Fuel node, you will typically not need to interact with the node's console.
Q: Is it possible to get any access to a demo, where i could get to know new versions better?
A: You can join our meet the expert sessions, or simply download Mirantis OpenStack 8.0.
Q: I'm interested in the OpenContrail plugin, but there are no updates for it since MOS6.1.
A: The latest Contrail plugin support is for the Kilo release, MOS 7.0. Mirantis is not directly working on an OpenContrail plugin. Community developers may be working on something, however.
Q: Can't Kubernetes be used for auto scaling of instances also?
A: Yes, the autoscaling functions of Kubernetes handles launches and removes instances within the OpenStack environment to support the required Kube minions supporting the workloads. This includes in the public GCE environment.
Q: Does Murano support Baremetal nodes for additional SW provisioning?
A: No. Murano does not support Ironic nodes out-of-the-box.
Q: How do you push to your subscribers the patches to your distribution? Are they notified in some automatic way?
A: Yes, registered users receive email notifications of technical bulletins. Also, they are posted on our support site: https://software.mirantis.com/mirantis-openstack-technical-bulletins/.
Q: Can you explain more about the Ceph H release used vs. the one used in Fuel 7? What about performance and erasure coding?
A: MOS 7 uses Ceph Firefly, while MOS 8 uses Ceph Hammer. We curate the individual features for hardening and stability. Please refer to the release notes for support features.
Q: Why are the the upgrade paths documented as "Experimental"? Can you guys break out the Upgrade path and documentation in a way that is easier to locate vs having it tucked into a Admin guide?
A: Upgrades have many aspects that need to be weighed to appropriately guide the processes, including environment customizations, number and types of plugins, features used, and your organizations SLAs. We provide a Professional Services offering to help you take these aspects into account and identify the best path to take.
For more information or to try it out yourself, please download Mirantis OpenStack 8.0.
