Shulou(Shulou.com)06/02 Report--

Storage replication is a new feature in Windows Server 2016. It is a native block-level replication technology on Windows Server, which can replicate disaster recovery scenarios such as partition-to-partition, stand-alone, extended cluster, and cross-cluster replication, and help organizations improve business continuity. Storage replication technology is asymmetric storage independent. In stand-alone and cluster architecture, each node can connect its own storage separately.

Main Technical Features of Storage replication Technology

Use SMB 3.1.1 communication protocol

Support for synchronous and asynchronous replication

Log disk and data disk are required for replication. The data is first written to the log disk, and then Commit data disk.

The smallest unit for each replication is Block.

Storage has no correlation, and the bottom layer of the node can be any storage structure

Support for fixed disks, the latest version of Server 2016 data center version, already supports thin standby disks

There is an active / standby relationship in the replication process. The primary replication partition can be read and written, while the standby replication partition cannot be read or written. Standby read-only mode is not supported.

Using security technologies such as packet signature and AES-128-GCM full-data encryption, the storage replication process uses Kerberos AES256 for all authentication between nodes.

In fact, storage replication is actually a disk filter driver that works on Windows Storage Stack, Partition Manager and Volume Manager. As we all know, a partition refers to a continuous storage area on a storage device, a volume refers to a logical collection of sectors, and the internal sector of a volume may have a partition, or multiple partitions, or different disks, while storage replication inserts another layer of logic between the partition to the volume and then exposes it to the volume. For the upper volume and application, you don't know that the bottom layer has done this. You can still use VSS technology, volume-level Bitlocker technology.

The working process of the storage can be further seen in the following figure. You can see that the storage replication is completed on the Partition Manager of the two nodes and under the Volume Manager.

Through this architecture, we can see the difference between storage replication and other replication technologies on Windows.

It is a block-based copy, and it is inserted in the area under the volume above the partition, so it does not know the concept of files, does not know that they are used, and does not know, like DFS, whether care files are being used.

It only cares about writing to IO and can be used for CSVFS,NTFS and ReFS

It is not based on checkpoints, but on continuous replication, so the increment of change is often much lower than that of snapshot-based products.

It runs on the partition layer, so all VSS snapshots created by Windows Server or backup software can be replicated

Lao Wang's comments:

Storage replication technology can be said to be a technology that the majority of ITpro has been looking forward to. In the previous version of Windows Server, there has been no good replication technology that can be used for virtualization, private cloud scenarios, DFS can only copy closed files, so many scenarios can not use it, the new storage replication technology can be said to be a bright spot, a very typical scenario, if an enterprise wants to implement a set of highly available cluster architecture This architecture may be remote, but there is no money to achieve device-level storage replication, so we have to use third-party products to achieve storage replication and disclose it to the cluster, such as Starwind or Datakeeper, and the host level to achieve software virtual storage replication, open to the cluster. Now with storage replication technology, we can directly use Microsoft native storage replication to achieve economical poor version of storage replication. The two points of storage and replication technology also lie in that its platform-independent and hardware-independent storage technology is implemented as a technology at the OS level. As long as you have Windows Server 2016 data center version, you can use this technology, so this is fun. It can be played in our local computer room, in private cloud, in public cloud, and in hybrid cloud scenarios. As long as you have OS, it's hard to say? Hardware-independent, storage replication does not restrict the underlying storage of nodes, which can be local SCSI/SATA,ISCSI,Share SAS and SAN. For stand-alone and cluster scenarios, you can also use ISCSI on one side and SAN on the other, as long as you ensure the data disk size and log disk size are the same.

Four usage scenarios of Storage replication Technology

Stand-alone partition pair partition

It can achieve partition-level replication for the partition above the stand-alone Server, and the replication protocol uses SMB. There are not many practical application scenarios, and disk mirroring, storage space and other technologies can be used instead.

Stand-alone to stand-alone

This scenario has some practical significance and can help two nodes in the same region or different regions to achieve block-based storage replication without clusters.

The technical features of stand-alone replication: manual failover, Powershell creation and maintenance, storage independence, and support for synchronous or asynchronous replication. Lao Wang will introduce the difference between the two later.

Extended cluster

Extended clustering means that a cluster with dual active storage is realized, avoiding the problem of storage in a single site and storage failure. Lao Wang mentioned this point in the multi-site and disaster recovery article before. Previously, we could only use third-party software or devices to achieve cluster storage replication. Now native storage replication and WSFC are perfectly integrated to achieve high availability + disaster recovery.

Extend the technical features of the cluster: GUI graphical interface management, only support synchronous replication (ensure cluster data consistency), storage independence, fully automatic failover

Cross-cluster replication

The realization of replication between two different Cluster should be rare in China. What Lao Wang can think of is probably a set of disaster recovery cluster. Maybe has a very important cluster system, which requires disaster recovery for the cluster itself and cluster data, so another cluster is built as it is. Usually, it is not external, and the data is copied to storage in real time, once the home cluster is broken or data is lost. The disaster recovery cluster starts immediately.

Technical features of cross-cluster replication: manual failover, support for synchronous or asynchronous replication, storage independence

Storage replication synchronous replication work process

Synchronization of working process of pattern diagram

Zero data loss

RPO1. Application writes data

two。 Write to the log disk and copy the log to the remote site

3. Remote site writes log

4. The remote site returned the write result

5. Replication engine response write complete, application IO ends

T&t1: refresh the log to the data disk later

Storage replication asynchronous replication work process

Pattern diagram step asynchronism

Near zero data loss

(depends on a variety of factors)

RPO1. Application writes data

two。 Data is written to the log disk

3. Replication engine response write complete, application IO ends

4. Copy logs to a remote site

5. Remote site writes log

6. The remote site returns a write completion confirmation message

T&t1: refresh the log disk data to the data disk later

Scenarios applicable to synchronous replication and asynchronous replication

Scenarios applicable to synchronous replication

Critical business application

Short-reach node (network delay destination replication group

Get-SRgroup

Displays the replication group information, which is mainly concerned with Replication Status. This attribute indicates that Relicating is being copied, and InitialBlockCopy indicates that it is still in the initial replication. If the word error appears, replication cannot be performed.

(Get-SRGroup). Replicas

Displays the replication status of each partition in the same replication group. Each replication group can replicate more than two partitions. This command can display the replication status of all partitions.

The log for storing replication can be seen through the event Viewer and is designed as two channels, Admin and Operational

Located in the application and service log-Windows-StorageReplica

Store replication performance counters. If there is a SCOM, you can use SCOM to collect performance counter pointers, or write MP packages for monitoring. You can also integrate OMS to display in the cloud.

Storage Replica Partition I/O Statistics

Storage Replica Statistics

Automatic failover is not supported in stand-alone or cross-cluster replication, so once the node is down, you need to manually switch replication, manually switch replication, in fact, the so-called reverse replication, we re-switch, by the available side of the primary node to provide read-write function.

Planned failover

Refresh the primary node replication group log to the data disk to prevent data loss

Sync-SRGroup-Name RG01-Force

Reverse copy command

Set-SRPartnership-NewSourceComputerName 16server2-SourceRGName RG2-DestinationComputerName 16server1-DestinationRGName RG1

After the reverse replication is completed, the current data is readable and writable in 16server2, and you can see the data we copied before.

If you are in an unplanned disaster recovery scenario and the primary server is suddenly powered off and down, you can use this command on the standby node to perform a forced failover. This command promotes the standby node to be the primary node. After the original primary node is restored, the direction can also be changed and copied back in reverse.

Set-SRPartnership-NewSourceComputerName 16server2-SourceRGName RG2-DestinationComputerName 16server1-DestinationRGName RG1

If the standby server goes down suddenly, it will not affect the replication. The primary server can still read and write normally, wait for the standby server to return to normal, and then synchronize the data.

Daily management operations of storage replication

Delete replication

# Delete the storage replication relationship and execute it on one of the nodes

Get-SRPartnership | Remove-SRPartnership-Force

# Delete a replication group, which needs to be performed on each node

16server1

16server2

The disk is released to each node separately, and the disk of each node can see the data.

16server1

16server2

Management authorization

By default, the local administrator of the storage replication server has the right to manage storage replication by delegating ordinary users without the need for local administrator privileges

Grant-SRDelegation-UserName oa\ mikewang

Restrict storage and replication Nic

By default, storage replication uses all available network cards for storage replication as far as possible. We can specify that the specified network card is used to complete the storage replication traffic.

Get-NetAdapter acquires the index information of each node network card

Create a storage replication network restriction policy

Set-SRNetworkConstraint-SourceComputerName 16server1-SourceRGName RG01-SourceNWInterface 11 13-DestinationComputerName 16server2-DestinationNWInterface 10 12-DestinationRGName RG02

Each node refreshes the SMB multichannel connection

Each node acquires SMB multichannel-dedicated links for storage replication

Limit the use of storage replication bandwidth

Create SMB bandwidth limit

Set-SmbBandwidthLimit-Category StorageReplication-BytesPerSecond 50MB

View SMB bandwidth limits

Get-SmbBandwidthLimit-Category StorageReplication

Remove SMB bandwidth limit

Remove-SmbBandwidthLimit-Category StorageReplication

Cannot configure replication again after deleting replication

Delete all orphaned Storage Replica partition databases and remount all partitions (stand-alone)

Clear-SRMetadata-AllPartitions

Delete all orphaned Storage Replica log data

Clear-SRMetadata-AllLogs

Delete all orphaned failover cluster configuration data

Clear-SRMetadata-AllConfiguration (clustering is a good move)

Delete individual replication group metadata

Clear-SRMetadata-Name RG01-Logs-Partition

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.