HP Helion OpenStack Cloud High Availability
When migrating business-critical applications to the cloud, you need to ensure their continuous operation and protect them from failures at the hardware or software level. When using classic server technologies for high availability, fault-tolerant clusters and redundancy of the main components of servers and storage systems are used, eliminating single points of failure. Let's see how high availability is implemented in the HP Helion OpenStack cloud.
High availability in the cloudy HP Helion OpenStack environment is implemented in three areas: high-availability cloud services, fault-tolerant cloud infrastructure, and tenant-supported cloud technologies, i.e. service providers providing cloud services to end users.
Let's begin the description of the applied solutions of high availability from loads of tenants of a cloud. Tenants themselves need to ensure the continued availability of applications that they deploy to the cloud for their customers, using virtual machines distributed across different availability zones and transferring network traffic to them using load balancers.
')
To improve the availability of applications, you can combine virtual machines into failover clusters. Similarly, the infrastructure on which the cloud itself is deployed (physical servers, storage systems, and network equipment) must always be accessible. Finally, the cloud provider needs to provide high availability of core cloud services and cloud APIs (Horizon, Keystone, Nova, Neutron, Cinder, Swift). For this purpose, replication of the main service controllers and their data is used, which provides fault tolerance of the control plane of the cloud. For example, for each installation, three separate upstream cloud controllers are deployed on separate physical servers, between which automatic replication occurs.
Also, software clustering and replication are used to ensure high availability of the database, message processing and web traffic proxy server, so if one upstream cloud controller fails, all its load will be redirected to another upstream cloud controller. Cluster Active-Active ensures continuous operation of the cloud and its users even in the event of a failure of one of the servers serving cloud services. Redundancy applies to Starter Swift object storage servers - for high availability, at least two servers are required. Swift software replicates data between these servers to back up Swift objects between two servers.
For Cinder block-based storage services based on HPVSA software-defined storage, high availability is provided through a three-node cluster connecting the servers on which VSAs are deployed. Also, when you deploy Cinder on an HP 3PAR disk array, you can use fault tolerance technologies developed for it.
High availability in HP Helion OpenStack can be implemented in several scenarios. High availability combined with load balancing is applied in four aspects of the cloud. First, the OpenStack Horizon user interface (UI) uses “sticky session load balancing” load balancing technology. Horizon is a stateless service. Using the session stickiness function, it is guaranteed that the resources of specific users will be tied only to the web session of the corresponding user.
The next area where high availability is implemented is the service APIs (Nova, Cinder, and so on). For them, OpenStack uses Round Robin model load balancing using HAProxy software. The next area where high availability is realized is the RabbitMQ messaging mechanism with the use of Master-Master mirroring. Finally, to ensure the high availability of the MySQL database, the Galera Cluster clustering software is used, which implements synchronous replication between nodes of a multi-node cluster. Galera Cluster protects MySQL from data loss, provides high availability and easy scaling as data increases.
Thus, to prevent failures in HP Helion OpenStack operation, “multi-echelon” protection is used based on multiple duplication of critical cloud components, combining both standard OpenStack technologies and HP's own development, and ensuring high availability of business applications deployed in the cloud.
Thank you for your attention, we are ready to answer your questions. Other publications about HP Helion, see the links in the block "Related publications" below.
High availability in the cloudy HP Helion OpenStack environment is implemented in three areas: high-availability cloud services, fault-tolerant cloud infrastructure, and tenant-supported cloud technologies, i.e. service providers providing cloud services to end users.
Let's begin the description of the applied solutions of high availability from loads of tenants of a cloud. Tenants themselves need to ensure the continued availability of applications that they deploy to the cloud for their customers, using virtual machines distributed across different availability zones and transferring network traffic to them using load balancers.
')
To improve the availability of applications, you can combine virtual machines into failover clusters. Similarly, the infrastructure on which the cloud itself is deployed (physical servers, storage systems, and network equipment) must always be accessible. Finally, the cloud provider needs to provide high availability of core cloud services and cloud APIs (Horizon, Keystone, Nova, Neutron, Cinder, Swift). For this purpose, replication of the main service controllers and their data is used, which provides fault tolerance of the control plane of the cloud. For example, for each installation, three separate upstream cloud controllers are deployed on separate physical servers, between which automatic replication occurs.
Also, software clustering and replication are used to ensure high availability of the database, message processing and web traffic proxy server, so if one upstream cloud controller fails, all its load will be redirected to another upstream cloud controller. Cluster Active-Active ensures continuous operation of the cloud and its users even in the event of a failure of one of the servers serving cloud services. Redundancy applies to Starter Swift object storage servers - for high availability, at least two servers are required. Swift software replicates data between these servers to back up Swift objects between two servers.
For Cinder block-based storage services based on HPVSA software-defined storage, high availability is provided through a three-node cluster connecting the servers on which VSAs are deployed. Also, when you deploy Cinder on an HP 3PAR disk array, you can use fault tolerance technologies developed for it.
High availability in HP Helion OpenStack can be implemented in several scenarios. High availability combined with load balancing is applied in four aspects of the cloud. First, the OpenStack Horizon user interface (UI) uses “sticky session load balancing” load balancing technology. Horizon is a stateless service. Using the session stickiness function, it is guaranteed that the resources of specific users will be tied only to the web session of the corresponding user.
The next area where high availability is implemented is the service APIs (Nova, Cinder, and so on). For them, OpenStack uses Round Robin model load balancing using HAProxy software. The next area where high availability is realized is the RabbitMQ messaging mechanism with the use of Master-Master mirroring. Finally, to ensure the high availability of the MySQL database, the Galera Cluster clustering software is used, which implements synchronous replication between nodes of a multi-node cluster. Galera Cluster protects MySQL from data loss, provides high availability and easy scaling as data increases.
Thus, to prevent failures in HP Helion OpenStack operation, “multi-echelon” protection is used based on multiple duplication of critical cloud components, combining both standard OpenStack technologies and HP's own development, and ensuring high availability of business applications deployed in the cloud.
Thank you for your attention, we are ready to answer your questions. Other publications about HP Helion, see the links in the block "Related publications" below.
Source: https://habr.com/ru/post/249879/
All Articles