How to unroll a hyper convergent system in the small company? TADetails
In a segment of large IT environments application of convergent infrastructure and the expert integrated systems — the ready-made solutions including servers, DWH, interconnects, means of virtualization and a management tool infrastructure is already for a long time approved. However, such systems badly are suitable for the small companies with limited resources. An exit - the hyper convergent systems in which the ideas of convergent infrastructure gained further development.
Content |
Convergent infrastructure
What to do if it is required to unroll IT infrastructure in the small company or branch? It is possible to take servers, DWH, switches, to set hypervisors and to turn virtual machines. Such approach has certain advantages, allowing to apply those products and solutions which in the best way approach customer-specific. However this approach bears the difficulties unacceptable for cases when the IT service of the company or branch does not incorporate a full range of specialists in servers, DWH, networks, virtualization and other infrastructure means since at creation of IT infrastructure using such approach (we will call it traditional) each of its subsystems is controlled and serviced separately by the specialist of the corresponding profile. Even in the presence in the company of all necessary specialists (that in a case with the small companies is rather an exception) there is a problem of lack of a uniform point of management that complicates processes of management of IT infrastructure and requires effective interaction between administrators of its subsystems.
At the same time in a segment of large IT environments application of convergent infrastructure and the expert integrated systems — the ready-made solutions including servers, DWH, interconnects, means of virtualization and a management tool infrastructure is already for a long time approved. The expert integrated systems are developed, made and implemented directly vendors (alliances of vendors) or system integrators. The customer receives the system already ready for deployment of its applications. The main distinctive features of such systems are, first, the unified environment of management including the management tools all subsystems integrated so that the processes which are required for accomplishment of tasks of management and mentioning different subsystems were executed automatically and in the necessary sequence. Such interaction between subsystems provided with a unified environment of management is called orchestration (engl. orchestration), and the environment of management or its component providing such interaction — an orkestrator (engl. orchestrator). Secondly, the uniform point of interaction with a support service of the supplier while at traditional approach it is required to interact with support services of each component of a system that considerably complicates the procedure of permission of complex incidents is inherent to the expert integrated systems. The uniform point of service also provides the centralized process of updating of a system and its components. Thirdly, such systems are usually constructed on the basis of the principles of convergent infrastructure: at such approach the equipment (servers, DWH, network switches, etc.) is presented in the form of resources (CPU time, memory, disk space, network ports), and the control of the equipment in general comes down to adding of these resources in this or that pool. To applications resources are appointed already from pools without binding to that equipment on which these resources physically are placed. It allows to manage flexibly resources and simplifies transactions in need of an equipment output from operation for service or replacement.
Unfortunately, the convergent expert integrated systems, being aimed first of all at an enterprise-segment, badly are suitable for the small companies with limited resources. Best of all these systems are suitable for deployment in large IT environments and data processing centers.
At the same time the small companies which problems at deployment of IT infrastructure are described in the first paragraph are capable to estimate the advantages of the integrated systems stated above too. In response to requirements of such customers recently the conducting suppliers of IT solutions began to offer the hyper convergent systems in which the ideas of convergent infrastructure gained further development. In the hyper convergent systems computing resources and storage system are united structurally in the small block of standard rekovy construct. Such block is the finished system ready to deployment of services of the customer and including server edges united in a virtualization cluster, and the virtualized disk storage integrating disk space of servers in one virtual DWH. A hyper convergent system is scaled by adding in the system of such blocks. The hyper convergent systems have all advantages of the integrated systems, but at the same time are available not only to big customers, but also rather small companies, and well are suitable for deployment of the main IT services that also allows to apply them to the organization of remote branches. In general the hyper convergent systems are a product of philosophy of software-defined DPC and can be example of implementation of software-defined approach to creation of IT infrastructure.
In practice
Example of a hyper convergent system is the HPE HyperConverged series. For example, HC250 model is constructed based on the chassis of high density of HPE Apollo 2000 with four servers edges and borrows 2U in a rack. The chassis supports installation of 24 disks of a small form factor, at the same time six disk compartments are physically connected to each server edge. In ConvergedSystem option servers are delivered with the preset hypervisors of ESXi 5.5 or 6.0 on one of which the virtual machine of management is placed. On each of nodes it is also located DWH component of a hyper convergent system — the StoreVirtual VSA virtual machine.
Installation
A system is delivered with fixture for rack-mounting. As it was noted earlier, a system occupies two units in a rack. System weight assembled exceeds 35 kg therefore for installation it is necessary to delete from it server edges, disks and power supplies, having remembered what servers what compartments occupied initially. It is recommended to mount a system by forces of two people. After installation of a system in a rack it is required to install its components in the chassis in the same compartments in which they were initially, to connect its power supplies two standard C13-C14 cables, then to connect SFP+ ports 10GbE of server edges to two switches of a factory area network. Optionally it is also possible to connect ports of management of iLO of server edges to the switch of network of management (for full-time job these ports are not required, but can be used in need of reset of a system in a factory status).
Here it is necessary to make the important remark. For system operation it is necessary to connect it in 10-gigabit network. Work in gigabit network though it is technically feasible (using the corresponding transceivers or with the order of 4-port adapters 1 GbE instead of two-port 10 GbE), can seriously reduce performance of a disk subsystem.
The switching circuit of a system is shown in the drawing below.
After installation and switching of a system it is necessary to be convinced that on it a power supply is given and server edges are included. After loading of servers (can take 10−15 minutes) a system is ready to deployment.
Deployment
As of March, 2016, means of OneView InstantOn is actively updated therefore accomplishment of the procedures described below can differ in the version of means relevant at the time of reading this text. We recommend to examine the current edition of the instruction for system deployment provided by the producer.
The procedure of system deployment is in detail described in the document HPE HyperConverged 250 Installation Guide available on the website of HPE. The description for a case of deployment of the new system consisting of one chassis and four servers is included below.
For system deployment it is necessary to connect the notebook or the PC of management to RJ-45 No. 2 port of server No. 1 (lower right if to look at a system behind) and to configure to network IP address 192.168.42.99/24 interface then under the RDP protocol to enter on the virtual machine of management (has address 192.168.42.100).
System deployment is made using means of OneView InstantOn. The master of deployment will be started automatically at login (also it can be started a label on a desktop and from the Start menu).
At start of the master of deployment it is necessary to accept conditions of user agreements, having noted the corresponding checkboxes and to click Next:
On the Health screen the current status of detection of nodes of a system and availability of the management of the vCenter server preset on VM will be displayed:
On each node of a system specially prepared version of a hypervisor of ESXi about which structure the agents providing interaction of a node with OneView InstantOn enter is set. It is necessary to wait until indicators become green color (it will not take more than one-two minutes) and to click Next.
On the following screen it is necessary to set the IP addresses of hosts of ESXi, interfaces of vMotion and interfaces of iSCSI. In total it will be required to select 23 addresses: 5 — for ESXi and vCenter, 4 — for vMotion and 14 — for iSCSI. All addresses can belong to one subnet, but also it is possible to set the subnet for each group of the addresses. We enter the addresses and we click Next:
On the Credentials screen it is necessary to enter a name and admin password of storage system StoreVirtual. These details will be used for management of storage. We click Next and we get on the Settings screen where we enter a name of storage (it is possible to leave offered by the master), the addresses of the DNS, NTP servers, and also the e-mail address on which a system will send notifications, and the SMTP server address, then we click Next. The Review Configuration screen will open:
On this screen the status of check of readiness of components of a system for deployment and correctness of the entered data is displayed. We are convinced that all points are marked in the green color, and we click Deploy. Process of system deployment during which to components of a system the specified IP addresses and other settings are set begins, hosts of ESXi of a system integrate in vSphere cluster, the cluster of VSA, etc. is created. System deployment takes about ten minutes. After deployment the page with the MAC addresses of the VSA nodes which are required for activation of licenses will open. Licenses are activated on the portal of support of HPE.
After deployment it is necessary to install licenses for vSphere and vCenter in the vSphere web console on the virtual machine of management.
Preparation of storage
After installation of licenses it is necessary to create storage on VSA. For this purpose it is necessary to select the necessary cluster from the vSphere web console, to go to the Manage tab and to click HP Management. The section of HP Management added by the preset plug-in HPE OneView for vCenter will open. This plug-in provides integration of vCenter with components of a system and allows to execute system management and monitoring of the vCenter console.
In the section of HP Management we go to the Storage tab and we click on the Actions menu. We select the item Create Datastore.
The master of creation of storage will open. We select location of the created storage — a cluster of vSphere and we click Next. On the following page of the master we select a storage pool (in this case there is only one pool), we specify the number of the created storages and their size. It is also possible to select the level of network RAID — for a configuration of a cluster of VSA from four nodes (i.e. for one HC250 system) the Network of RAID 10 is recommended. We click Next. On the next step we set a name of storage and we click Next. On the next step validation of the entered data is made. It is necessary to wait for its end and to click Next, then — Finish. The storage will be created within several seconds and at once will become available to use. Now a system is ready to work. Deployment of virtual machines and other daily transactions do not differ from like those vSphere in normal versions.
Reset of a system
If something went wrong, a system can be reset in a factory status. All data at the same time will be lost.
For reset of a system it is necessary to download tool kit from the website of HPE. Scripts for deployment and an image of the virtual machine of management enter it the customized ESXi distribution kit with the integrated image of the VSA virtual machine.
In general the procedure of reset of a system looks as follows: at first the installation USB carrier by record of an image by the dd tool or similar is created, then scripts for deployment of specific model of a system then each of servers is loaded from the carrier (here it is recommended to use the module of management of iLO) register in it and the hypervisor is established. Then on server No. 1 the image of the virtual machine of management then a system is ready to the automated deployment is developed.
In detail the procedure of recovery is described in the document HPE HyperConverged 250 User Guide available on the website of HPE.
Material is prepared with assistance of HPE and Lanit companies
| The site content is translated by machine translation software powered by PROMT. The machine-translated articles are not always perfect and may contain errors in vocabulary, syntax or grammar. Read original article If you find inaccuracies or errors in the results of machine translation, please write to editor@tadviser.ru. We will make every effort to correct them as soon as possible. |