Achieve mission-critical standards of data availability
Contents
1. By Management Group Y Show Details E aD LHN_Mmgt_Group Disk Protection Using RAID Disk RAID Consistency Cluster Level Data Protection Volume Level Data Protection Volume Access Network Bonding Network Bond Consistency Network Flow Control Consistency Network Frame Size Consistency System alarms System alarms are visible from the CMC and display all warnings and critical events for all management groups that the user is logged into System alarms serve to alert the system administrator of any possible misconfigurations errors in the system or hardware faults To maintain high uptime of HP StoreVirtual Storage the alarm list should be checked regularly and should any alarms be present corrective action should be taken to clear the alarm list Figure 5 System alarms showing items that need attention f amp Alarms 7 Status Normal Fitter Alarms Severity Y Date Time Event Message Type A Warning 8 31 11 4 21 58 PM MDT E02000202 EID_ Email Configuration S Configuration Check A Warning 8 31 11 4 22 10 PM MDT E02000202 EID_ Email Configuration S Configuration Check A Warning 8 31 11 4 22 10 PM MDT E01090101 EID_ Power Supply 1 Stat PowerSupply A Warning 8 31 11 4 22 25 PM MDT E02000202 EID_ Email Configuration S Configuration Check A Warning 8 3141 4 22 25 PM MDT E02000202 EID_ Email Configuration S Configuration Check A Warning 8 31 11 4 222. to create a scalable storage pool The storage cluster aggregates the critical components of a number of identical storage systems into a single pool of resources This storage cluster accepts and responds to iSCSI requests as a single logical system In HP StoreVirtual Storage all physical capacity is aggregated and is available to all volumes created on the storage cluster When more storage capacity or performance is needed additional HP StoreVirtual Storage nodes can be added to the storage cluster online the storage seamlessly non disruptively reorganizes its storage to incorporate the new node HP StoreVirtual Storage is inherently highly available and features integrated disaster recovery features HP StoreVirtual Storage also scales capacity and performance linearly as storage nodes are added to the cluster providing a predictable growth strategy as new storage requirements need to be satisfied Finally management of an HP StoreVirtual Storage cluster is straightforward and simple to learn regardless of the number of nodes in a cluster 1Source HP Storage quality review weekly data August 2012 Advantages of HP StoreVirtual Storage e Availability HP StoreVirtual Storage can sustain multiple concurrent failures and still keep data online and accessible Most storage architectures use redundant components to ensure HA Yet 70 percent of unplanned downtime is not caused by component failure but human error Typically redundant co
3. 10 2 Four copies of data replicated between nodes Protection against any three nodes going offline Protection from site failure Protection against any seven drives failing in the cluster 5 Single parity distributed across all nodes in Protection against any node going offline the cluster Protection against any three drives failing in the cluster 6 Dual parity distributed across all nodes in Protection against any two nodes going offline the cluster Protection against any five drives failing in the cluster When configured as a multisite SAN Figure 1 Various levels of Network RAID Granular high availability Storage Complete site protection via multisite SAN configuration The multisite SAN feature leverages the core functionality of Network RAID 10 and 10 1 One of the advantages of Network RAID 10 is that there are two complete copies of the data in each volume and each copy is guaranteed to be stored on a different node Due to Network RAID 10 s data layout one of the net effects of this is that every other node can go offline at the same time In other words half the HP StoreVirtual cluster can go offline and the volumes and associated applications will stay online with no intervention from the storage administrator This level of availability is not found anywhere else in the storage industry and while this is a powerful feature its value is improved even more when the HP StoreVirtual cluster is split acro
4. 29 PM MDT E02000202 EID_ Email Configuration S Configuration Check Warning 9 41 8 07 58 PM MDT E010A0102 EID Network Interface M Network Online upgrades with Upgrade Advisor Online upgrades give storage administrators the ability to apply the latest software and firmware to their systems without bringing the storage cluster down for maintenance There can be cases where improvements to the HP StoreVirtual Storage nodes are available via firmware upgrades or upgrades to the LeftHand OS previous versions have been referred to as SAN iQ Some of these improvements are available via major version upgrades such as from SAN iQ 9 0 to SAN iQ 9 5 Other upgrades may be available as smaller individual pieces of software or as a group of software known as a patch Set Regardless of the type of upgrade the Upgrade Advisor in the CMC automatically checks the current installed software levels on the nodes in the cluster against generally available software releases published by HP and alerts the system administrator when an upgrade is available Additionally the Upgrade Advisor provides the option to implement the software upgrades and provides a checklist of dependencies that may need to be remedied elsewhere in the environment for example ensuring that the HP StoreVirtual DSM for MPIO software is compatible with the upgrades for the storage nodes As these software upgrades are made available to improve availability of systems
5. Advanced tab and choosing a new cluster from the cluster drop down box The blocks that make up the volume on the original cluster will begin to migrate to the new cluster and the LeftHand OS will automatically redirect and proxy requests for blocks to the proper cluster as the data migration is underway When the migration is complete the iSCSI sessions to the new cluster from the host are automatically restored assuming the new cluster s virtual IP address has been added to the iSCSI configuration of the host server A typical use case for Peer Motion could be a volume that contains data for an application that has increasing performance needs If the volume started out on an MDL SAS cluster a storage administrator could use Peer Motion to move the volume to a SAS based cluster If the volume is on a SAS cluster the storage administrator could choose to add more nodes to the cluster to provide more performance for the volume or they could choose to move the volume to an even higher performing tier such as an SSD based cluster Figure 8 Peer Motion for volumes moving volume to higher tier of storage Basic Advanced Cluster LHN_SAS_ Cluster z LHN_SAS_Cluster Data Protection L 1H at rotecuon eve AN SSD_Cluster Type Primary O Remote Provisioning Full O Thin 10 Peer Motion on HP StoreVirtual Storage Cluster swap The virtualization of storage within an HP StoreVirtual cluster means th
6. Fault Tolerance on HP LeftHand Multisite SAN cluster Technical white paper http h20195 www2 hp com v2 GetPDF aspx 4AA0 4385ENW pdf HP P4000 best practice guide for Hyper V Technical white paper http h20195 www2 hp com v2 GetPDF aspx 4AA4 0956ENW pdf HP LeftHand Storage with VMware vSphere Technical white paper http h20195 www2 hp com v2 GetPDF aspx 4AA3 6918ENW pdf Peer Motion on HP LeftHand Storage Technical white paper http h20195 www2 hp com v2 GetPDF aspx 4AA4 0304ENW pdf For more information To know more about how you can drive your data center to achieve new industry standards of data availability visit hp com go StoreVirtual Get connected hp com go getconnected Current HP driver support and security alerts delivered directly to your desktop Copyright 2012 Hewlett Packard Development Company L P The information contained herein is subject to change without notice The only warranties for HP products and services are set forth in the express warranty statements accompanying such products and services Nothing herein should be construed as constituting an additional warranty HP shall not be liable for technical or editorial errors or omissions contained herein 4AA4 2922ENW Created November 2012
7. availability Network RAID 5 6 are viable for volumes that have less stringent performance requirements Once you have ensured that all volumes are protected by Network RAID simply proceed with the maintenance one node at a time When the maintenance for that node is finished the node comes back online and the data is resynchronized When the resynchronization is complete the next node in the cluster can undergo maintenance Redundant hot swappable components in each node Every HP StoreVirtual Storage node in a cluster has hardware that is protected by redundant hot swappable components for cooling and power This hardware redundancy protects a node from the most common hardware failures so that a hardware failure does not cause disruption to the HP StoreVirtual Storage cluster Additionally the physical disk in each node is protected by enterprise level disk RAID which can be configured by the storage administrator during set up of the storage node to meet the performance capacity and availability requirements The best practice and default is hardware RAID 5 Protected non volatile write cache Every write that is sent from the application to HP StoreVirtual Storage is guaranteed to be protected by persistent cache before the write is acknowledged back to the host This ensures that your data is protected in the event of an unexpected power outage in the middle of operations When power is returned to the storage system the writes that ar
8. managers at each site and the FOM running at a third site This allows quorum to be maintained in the event of a site becoming unavailable 4 When using a failover manager FOM ensure that the FOM is not installed on a datastore that is on HP StoreVirtual Storage The datastore for the FOM must be on non HP StoreVirtual Storage if the FOM is on HP StoreVirtual Storage a circular reference exists and the system is no longer in a valid HA configuration For more information about managers and multisite configuration please read the Product User Manual and Multisite SAN User Manual available at hp com go LeftHandDownloads Configure cluster HP StoreVirtual Storage clusters can consist of one to 16 nodes per HP best practices While a single node cluster can be configured it lacks the ability to have data protection and availability outside of the single node as Network RAID requires a minimum of two nodes in a cluster one copy of the data per node For this reason a single node cluster does not meet the HA requirements as described in this document It is strongly recommended that an HP StoreVirtual cluster comprise of two or more nodes to leverage features such as Network RAID to improve data availability Create volume Network RAID is a feature available on a per volume basis that replicates the contents of the volume across different nodes in a cluster transparently to the host server and application Network RAID protects against many
9. Technical white paper Achieve mission critical standards of data availability HP StoreVirtual Storage Table of contents Executive summary HP StoreVirtual architecture Advantages of HP StoreVirtual Storage HP StoreVirtual HA features Network RAID Complete site protection via multisite SAN configuration VMware vSphere Metro Storage Cluster Best practices analyzer System alarms Online upgrades with Upgrade Advisor Quorum and managers No downtime for maintenance Redundant hot swappable components in each node Protected non volatile write cache Background error checking correction Online volume configuration changes Peer Motion on HP StoreVirtual Storage Volume migration Peer Motion on HP StoreVirtual Storage Cluster swap Insight remote support Conclusion Appendix 1 Configuring an HP StoreVirtual cluster to enhance HA features Appendix 2 Checklist for optimal availability Appendix 3 Useful links For more information WOWOWOUWO MA ONDA AUWW WN N 13 15 16 16 Executive summary The rapid adoption of virtualization technologies as well as the rapid growth in data requires a shared storage system that is available at all times While 99 99 percent uptime annual downtime of 52 minutes used to be the accepted norm for most organizations the dependence on shared storage has changed the requirement so that 5nines Sminutes 99 999 percent availability 5 minutes of downtime per year is the new standard for data avai
10. at the rules about data being tied to physical hardware resources no longer applies This virtualization allows volumes to be moved dynamically between different physical hardware clusters and also allows for a feature called cluster swap the ability to remove existing storage nodes from a cluster and replace them with new storage nodes online with no loss of data or data availability In one operation data from the old storage nodes is moved to the new storage nodes and all IO is properly directed to the correct node Upgrading to newer faster or larger storage nodes does not require any downtime providing a clear well defined strategy for future expansion and growth As an example a customer might start out with a cluster of 8 drive systems As the customer adds more applications and workload to the cluster they could reach the performance or capacity limit of the nodes They could easily migrate to nodes with 12 or more drives to increase capacity and performance without having to bring any applications offline Figure 9 Swap storage systems through the CMC HE LHN_SAS_Cluste LHN_SAS Cluster New Volume Edit Cluster gt Edit Cluster Properties Delete Cluster Add Storage Systems Help Remove Storage Systems Exchange Storage Systems Swap Storage Systems Reorder Storage Systems Edit Virtual IP Addresses Edit iSNS Servers Figure 10 Select the storage nodes to swap Fe Swap Clu
11. d any volumes that are configured with Network RAID 0 are not highly available Network RAID is configurable on a per volume basis and can be changed to a different Network RAID level on the fly ensuring the highest levels of availability while still providing the flexibility to adapt to changing requirements and workloads The beauty of Network RAID is that the protection and performance happen on the storage system level meaning that the host and application are unaware of the extra protection that is happening in the storage system In the event of a node going offline for maintenance upgrade and so on the host and application are unaware of any changes that are happening on the storage level and continue to run uninterrupted Table 1 Summary of different Network RAID levels Network RAID level Description Protection level 0 One copy of the data with no replication between Network RAID 0 is not considered part of an nodes ina storage cluster HA solution 10 Two copies of data replicated between nodes ina Protection against any node going offline storage cluster Protection from site failure Protection against any three drives failing in the cluster half of all HP StoreVirtual 4000 systems can go offline and yet the data will remain online 10 1 Three copies of data replicated between nodes Protection against any two nodes going offline Protection from site failure Protection against any five drives failing in the cluster
12. e in the protected cache get destaged to hard drives in the storage node Background error checking correction There are several features that continually monitor the data on the HP StoreVirtual Storage cluster automatically without administrator intervention such as e Disk controller scrubbing The HP Smart Array RAID controller card continually scans and scrubs the data in the disk RAID set looking for bad blocks and repairing them from parity calculations should the need arise e Self healing storage The LeftHand OS also has a low impact background operation which continually reads blocks of data and in the event that a bad block is found the block is marked as bad and the data that was on that block is reconstructed via the parity or copy from Network RAID and written to a known good sector e LeftHand OS read error recovery In the event that a bad block is discovered during a read operation the block is reconstructed from Network RAID For example if a drive fails and is replaced the disk RAID set is rebuilt using parity from the remaining drives in the RAID set If a bad block is found during the RAID rebuild it could cause the rebuild to fail LeftHand OS will see that the block is bad and pull a good copy of the block from the Network RAID copy or reconstruct it in the case of Network RAID 5 6 These features work together to enable that the data on HP StoreVirtual Storage is ina known good state and readily accessible to the ho
13. ed dramatically When a new management group is created the LeftHand OS automatically configures and starts the proper number of managers for that management group However as nodes are added and removed from management groups the management group may no longer have the proper number of managers 13 To ensure the proper number of managers are running in a management group 1 From the CMC check the best practices summary under the configuration summary section in the left column of the CMC 2 If there are an improper number of managers running the best practices summary will inform the user and suggest the proper number of managers a If too many managers are running right click on a node running a manager and select Stop Manager You will be asked to confirm this decision Repeat on different nodes until the proper number of managers is running b If there are not enough managers running right click on a node that is not running a manager and select Start Manager c If this is a two node cluster an FOM is highly recommended If an FOM cannot be used a Virtual Manager must be added to the Management Group to provide availability If there are different generations of nodes in the management group it is recommended to run managers on the larger or newer nodes 3 Ina multisite cluster configuration where a management group and or cluster spans two different physical locations best practices are to have an equal number of
14. fault scenarios including multiple disk failure individual node failure and even total site failure when the system is in a multisite configuration It is highly recommended to enable Network RAID 10 or higher on all volumes in a cluster to provide high uptime By default a volume is created with Network RAID 10 If a volume is configured as Network RAID 0 no redundancy across nodes the Network RAID level can be changed online with no disruption to service to the host server or application by doing the following 1 Select the volume in question right click and select Edit Volume 2 Click on the Advanced tab in the box that appears 3 Under Data Protection Level choose the proper Network RAID level 4 Click Ok The volume will begin to create the proper redundancy based on the Network RAID level selected Note You will not be allowed to select a Network RAID level that does not have the proper number of nodes present For example if you select Network RAID 10 2 four copies of data and only have three nodes you will not be able to choose that Network RAID level 14 Appendix 2 Checklist for optimal availability The BPA and System Alerts will highlight most of the configurable options that can impact availability It is recommended to log in to the CMC on a regular basis and check the BPA and Alerts for any actionable items that are being reported and address any of these concerns in a prompt manner To enab
15. for issues that storage administrators can control it is recommended that any available upgrades found by the Upgrade Advisor be reviewed and implemented at the earliest Figure 6 Upgrade Advisor showing that upgrades are available at SAN Status P Status Page Summary A Upgrades p Getting Started 7 aa A graye Summa Download Started 6 12 12 11 18 PM N_Mmgt_Group E E gt Events Download Finished 6 12 12 11 18 PM iis Servers 0 Click here to stay at current software version s Administration Sites All filles have been successfully downloaded HH LHN_SAS _Cluster fz Centralized Management Console Upgrades Available Advanced A Upgrades Available Advanced LHN_Mmgt_Group Upgrade the CMC first P4000 Application Aware Snapshot Manager 10 0 0 1222 should be currently installed P4000 DSM for MPIO 10 0 0 1224 should be currently installed P4000 Command Line Interface CLI 10 0 0 1301 should be currently installed Quorum and managers Many checks need to be in place for a distributed architecture to function properly Like other clustered architectures HP StoreVirtual solves this potential issue with the concept of quorum A manager is a specialized process that runs on one or more of the storage nodes in a management group Managers use a voting algorithm to coordinate storage system behavior In this voting algorithm a strict
16. instructions can be found here http h18000 www1 hp com products servers management insight remote support index html Figure 11 Sample HP Insight Remote Support configuration Tools Deploy Configure gt Diagnose Reports Tasks amp Logs Options Help v All Systems System s ka Quick Launch View as table v I Select ail Systems itself Summary amp 2 Critical W10 Major 0 Minor 119Normal E1 Disabled 12 Unknown Totat 144 g System Address Product Name C O example Server 255 255 255 255 S OD example Server 255 255 255 255 O0 0 example Server 255 255 255 255 ia64 HP UX B 11 23 Oo O example Server 255 255 255 255 O0 0 example Enclosure 255 255 255 255 Power Enclosure LJ r o example Server 255 255 255 255 9000 800 HP UX 6 11 11 O0 example Server 255 255 255 255 9000 800 HP UX B 11 23 DO example Server 255 255 255 255 C e onclusion Now more than ever shared storage is a critical piece of an organization s IT infrastructure Server virtualization and the exponential growth of data require not only the scalability and performance of shared storage systems but require that the storage be available at all times Snines Sminutes of availability used to be a goal for storage it is now a requirement for organizations HP StoreVirtual SANs can inherently provide Snines 5minutes or greater of availability to all users regardle
17. lability HP conducts standard quality reviews of all reported instances of data unavailability and data loss for all storage systems covered by a support agreement This quality review applies to all HP enterprise class storage servers and networking products In this review process HP tabulates data unavailability hours as reported by customers Given this and the number of systems under warranty we can estimate field availability Field availability is defined as the availability of a host or server to access data that is on the HP StoreVirtual Storage cluster If the host or server cannot access the data due to connectivity issues or if the data is not in a good state the storage cluster is considered unavailable Over the last two years HP has been able to determine that HP StoreVirtual Storage delivers Snines Sminutes or greater availability in the field when configured according to best practices The purpose of this paper is to describe the best practices and high availability HA features inherent to HP StoreVirtual Storage This paper provides e Guidance on their benefits and how they operate e Instructions on how to utilize these features to improve availability The appendices in this document provide guidance on how to enable those features which require manual configuration HP StoreVirtual architecture HP StoreVirtual 4000 Storage is an iSCSI based storage system that uses storage clustering a form of scale out storage
18. le that the system meets HA requirements the following conditions should be met BPA should be all green Alarms list should be clear no alarms present Software and firmware should be on latest versions no software available via Upgrade Advisor Additionally the following items can be checked on a regular basis to improve system uptime Network bond Make sure each NIC in the bond is active An alarm will be present if there is an issue with the networking Correct number of managers On the BPA Systems Running Managers should be green If not the BPA will provide guidance on the proper remedy Hardware RAID status lIn the BPA the Disk RAID consistency object should be green If there is an issue with the disk RAID consistency the BPA will highlight it and it will appear in the alarms section of the CMC as well Network RAID on volumes In the BPA the Volume level data protection item should be green indicating that all volumes being monitored are protected with a Network RAID level higher than 0 If any volumes that are being monitored by the CMC are at Network RAID level 0 the BPA will show an alert icon and provide guidance on which volumes are not in an HA configuration Alerts configured Ensure that email settings and SNMP settings are correct for whichever notification mechanism you choose to implement In the event of an alarm or incident that could impact availability it is important that the HP Sto
19. lover manager Site Rack2 Site Rack 1 Single logical server cluster Server Single logical storage cluster volume with Network RAID 10 D Storage Storage Failover manager Figure 3 Multisite SAN with site failure datastores stay online and virtual machines move to online site Site Rack2 Server cluster Storage Failover manager VMware vSphere Metro Storage Cluster The HP StoreVirtual Storage solution was the first iSCSI storage offering to receive the vSphere Metro Storage Cluster vMSC certification vMSC is a VMware certified configuration for stretched storage cluster configurations and any storage product that has vMSC certification must pass a series of tests issued by VMware In other words a vMSC certified configuration is designed to maintain data availability beyond a single physical or logical site The vMSC certification validates that the HP StoreVirtual Multisite SAN configuration is fully compatible with the VMware HA features over geographically dispersed sites This validation ensures that there are no technical obstacles or incompatibilities to overcome when deploying a multisite VMware solution vital for maintaining high availability Achieving the first iSCSI certification speaks of the close collaboration between VMware and HP for HP StoreVirtual Storage Best practices analyzer The majority of availability issues with HP StoreVirtual Storage can be resol
20. majority of managers a quorum must be running and communicating with each other in order for the LeftHand OS software to function The BPA provides guidance on whether the proper number of managers is running and it should be checked on a regular basis No downtime for maintenance Any volume protected by a Network RAID level higher than 0 can withstand any storage node going offline whether for maintenance or due to unexpected circumstances Because of this maintenance can be done at any time ina live environment without having to bring down applications or hosts Simply choose a storage node to carry out the maintenance on and perform the maintenance Even if that specific node becomes unavailable due to a software installation a hardware installation that requires a power shut down or a maintenance reboot the volumes that are protected with Network RAID higher than 0 will remain available This means that there is no longer a requirement for a storage maintenance window simply carry out maintenance on one node at a time and Network RAID will allow HP StoreVirtual Storage to continue to serve data If one or more nodes in an HP StoreVirtual cluster or management group need maintenance and there is a requirement for volumes to remain online each volume must be protected by a Network RAID level higher than 0 Network RAID 10 is the best practice that is the recommended Network RAID level as it offers the best combination of performance and
21. mponents such as two power supplies two fans and two NICs ports don t protect against power HVAC network building or site issues As long as HA is conceived only as having multiple redundant components inside the array all failure scenarios occurring outside of the array render these arrays unavailable HP StoreVirtual provides levels of data availability beyond what traditional dual controller storage and hardware RAID provide e Cost and simplicity Buy only what you need today and grow your storage non disruptively in the future Most storage architectures require you to plan for growth and over purchase initially With HP StoreVirtual you can start with a small capacity and scale to hundreds of terabytes all managed from a single intuitive user interface e Scalable performance Because the resources of all HP StoreVirtual nodes are aggregated in a storage cluster capacity and performance of the cluster increases each time a storage node is added This avoids expensive and complicated controller upgrades Adding HP StoreVirtual nodes into a cluster is done online without disruption transparent to the applications and servers This avoids downtime and complex coordination with server application and network teams HP StoreVirtual HA features Network RAID Network RAID is a unique feature of HP StoreVirtual Storage that offers protection above and beyond typical hardware redundancy Hardware RAID 10 is understood as mirrored and s
22. reVirtual Storage system is able to report the incident to the correct administrator or storage team Email SMTP configured tested for alerting lIf email SMTP is not configured properly an alarm will remain present in the CMC with a message similar to The Email configuration status is Unconfigured To be in an HA configuration all alarms need to be cleared from the CMC VIP load balancing enabled lf VIP load balancing is enabled on all defined servers the BPA will be green for the item Volume Access If any of the defined servers do not have VIP load balancing enabled the BPA will show an alert and list the servers that need attention 15 Appendix 3 Useful links The Architectural Advantages of HP StoreVirtual Storage US English http h20195 www2 hp com v2 GetPDF aspx 4AA3 0365ENW pdf The important role of storage for success in server virtualization http h20195 www2 hp com v2 GetPDF aspx 4AA2 5423ENW pdf The important role of storage for success in business continuance http h20195 www2 hp com v2 GetPDF aspx 4AA2 5417ENW pdf Implementing VMware vSphere Metro Storage Cluster with HP LeftHand Multisite storage Technical white paper http h20195 www2 hp com V2 GetPDF aspx 4AA4 0955ENW pdf HP P4000 Multisite HA DR Solution Pack User Guide http bizsupport1 austin hp com bc docs support SupportManual c03041871 c03041871 pdf Best practices for deploying VMware vSphere 5 with VMware High Availability and
23. ss different locations The cluster is still logically represented as a single pool of storage connected by the same networking protocols as ina single site As an example in a single data center one half of the cluster can be on a different power circuit than the other half The storage nodes continue to operate as a single cluster through the same standard networking protocols as they would if all nodes were ina single power circuit When volumes are configured with Network RAID 10 should a power breaker trip the other half of the cluster remains online with a complete copy of the data in the volumes The applications that are dependent on this data still have access to it and so the applications continue to be online Taking it one step further with two data centers if half of the cluster is in each of the data centers should an entire data center go offline for some reason power cooling or natural disaster half of the cluster is still running with a complete copy of the data and no user intervention is required to allow the HP StoreVirtual Storage cluster to continue to serve data This configuration is known as a multisite cluster When configured with Network RAID 10 1 a multisite cluster can span three data centers one copy of the volume at each data center providing a premium level of site redundancy and availability Network RAID 5 and 6 are not supported in a multisite cluster Figure 2 Multisite SAN configuration with fai
24. ss of the scale of the cluster using features built into the product with no additional charge Additionally many of these features are easy to use or require no user intervention HP StoreVirtual is the only storage product to offer continuous availability in the event of disk node data center or site failure with no intervention from the system administrator By leveraging the technologies and features described in this paper all HP StoreVirtual customers enable that their storage provides continuous data availability to keep their applications and organization online even during the most extenuating of circumstances HP strongly encourages all customers to follow the suggestions as described in this document 12 Appendix 1 Configuring an HP StoreVirtual cluster to enhance HA features The following section is designed to provide a step by step reference for configuring a new installation to take advantage of the HA features in HP StoreVirtual Storage While this document is intended to be as accurate as possible feature enhancements may mean that the guidance in this document can change between major release versions It is highly recommended to check the official documentation for the product or feature in use prior to implementing any changes HA on Storage nodes To improve availability for each node it is recommended to leverage network bonding on the network interfaces as well as choosing a disk RAID level that is best suited for
25. st servers and applications Online volume configuration changes Because the underlying physical storage is virtualized there are no hard rules about where the data can be stored or how volumes can be configured This virtualization of storage allows administrators to change just about every feature of an HP StoreVirtual volume including size Network RAID level whether the volume is thin provisioned and so on As many of these tasks are simply accounting changes meaning a count of how many blocks are assigned to a particular volume they can be done instantly with no impact to the system performance Unlike other storage systems changing the properties of a volume requires no downtime for either the volume itself or the host accessing the volume giving storage administrators the flexibility to adjust to changing requirements Figure 7 Sample volume properties that can be changed online fe Edit Yolume Ed Basic Advanced Cluster LHN_SAS_Cluster x Data Protection Level Network RAID 10 2 Way Mirror z Type Primary O Remote Provisioning Full O Thin Peer Motion on HP StoreVirtual Storage Volume migration Peer Motion on HP StoreVirtual Storage allows a system administrator to move an HP StoreVirtual volume from one cluster to another online without having to reconfigure the host or applications This is done by simply editing the properties of a volume selecting the
26. ster Storage Systems LX Swap storage systems assigned to the cluster with storage systems that are available For each storage system that you want to swap out select an available storage system in the column on the right Storage systems assigned to cluster Storage systems to be assigned LeftHand Node 1 LeftHand Node 4 bA LeftHand Node 2 LeftHand Node 3 LeftHand Node 5 Sep SHEE a 11 Insight remote support Insight remote support is a software tool that remotely monitors troubleshoots and facilitates problem resolution for your hardware infrastructure servers storage and network devices HP Insight Remote Support is an HP remote support solution for businesses Continuously monitoring your environment 24x7 this solution proactively alerts you about potential problems and troubleshoots for you when problems occur With accurate diagnostics information is sent to HP support staff or authorized partners who will determine if the problem can be resolved immediately or if an onsite visit is required It is highly recommended that HP Insight Remote Support be installed and configured to monitor all HP StoreVirtual Storage nodes This provides both proactive and reactive monitoring and problem resolution to speed up problem resolution leading to increased uptime and availability More information on HP Insight Remote Support as well as links to software downloads and installation
27. the environment To enable network bonding do the following 1 From the console connection on the node enter the proper node name IP address subnet mask and default gateway if applicable for each of the HP StoreVirtual nodes Install the CMC and find the nodes by IP address For each node a Enable the second NIC and make any changes to flow control and or jumbo frames b Select the two NICs and choose to create a bond c Choose the bonding type appropriate for the networking design and assign the proper IP subnet and gateway if applicable Note Any changes to flow control or frame size must be done on each NIC prior to creating the bond Flow control and frame size cannot be changed once a network bond is created The CMC can automatically check for available firmware and software updates To check if the HP StoreVirtual Storage nodes have available updates simply log in to each node by double clicking on it from the CMC Once logged in the CMC will check firmware and software versions of the nodes against a list of available versions and prompt the administrator to do the upgrade if newer firmware or software is available Configure management group As discussed earlier managers play a pivotal role in the scalability and availability of HP StoreVirtual Storage By following best practices around the proper number of managers in a management group and understanding the rules about quorum availability of the system can be increas
28. triped blocks within a disk array group Similarly Network RAID levels 10 10 1 and 10 2 synchronously replicate blocks of a provisioned volume across nodes in an HP StoreVirtual cluster transparent to the operating system or host connected to the storage In other words the host still sees a single logical volume as presented by the cluster it is unaware of the extra level of protection happening in the cluster This synchronous replication allows any single node in the cluster to go offline for any reason without affecting access to volumes protected by Network RAID By its definition a Network RAID 10 volume consumes capacity at twice the rate of a Network RAID 0 volume Network RAID 5 and 6 provide parity based replication of data at the back end providing protection against a single or double offline node while consuming just a portion of the capacity of Network RAID 10 10 1 or 10 2 While Network RAID 10 based replication provides multiple complete copies of a volume Network RAID 5 and 6 use parity to reconstruct data in the event that a node goes offline Network RAID 10 is comparable to disk RAID 10 while Network RAID 5 and 6 are comparable to disk RAID 5 and 6 respectively As there is a performance impact associated with the parity calculation it is recommended that Network RAID 5 and 6 be used for file shares or archives or other volumes that are not write intensive Network RAID O provides no redundancy between nodes in a cluster an
29. ved by following established best practices To provide guidance on compliance with recommended best practices the best practices analyzer BPA is built into the centralized management console CMC The BPA compares all configuration decisions with best practices and highlights configuration issues such as improper NIC teaming or lack of it checking data protection with Network RAID proper load balancing across nodes in the cluster and many more It is strongly recommended that the BPA be checked on a regular schedule especially immediately after making any changes to the storage cluster such as but not limited to adding or removing nodes creating new volumes or making changes to the networking configuration The BPA can be found on the SAN status page part of the CMC that provides a single high level view of alarms cluster usage available upgrades and more for all nodes that are being managed from that CMC Note that the SAN status page and BPA will not provide guidance on misconfigured options For example the SAN status page will alert the system administrator if the wrong number of managers is running ina management group However it will not alert the system administrator if the failover manager FOM is misconfigured such as the FOM virtual machine running on the same HP StoreVirtual Storage cluster that it is helping to manage Figure 4 BPA showing best practices being followed a F Best Practice Summary 0
Download Pdf Manuals
Related Search