伴随着vSphere 6.7的发布，vSAN也一同更新。看似只是小版本号多了0.1，但新功能上的变化可是真不少。
下面我会贴出一篇国外网站（https://www.vladan.fr/vmware-vsan-6-7-announced）上的原文，同时加入了自己的一些理解。我不算这方面专家，如有写错之处望大家多指正：）
VMware vSAN 6.7 What’s New?
Adaptive Resync（自适应重同步）
–
dynamic management of resynchronization and VM traffic. While the
previous release of VMware vSAN allowed putting a manual threshold on
the resync operations and also to split large components during
redistribution, the vSAN 6.7 goes much further and allows adaptive
resync.
Adaptive resync will ensure that a fair allocation of
resync resources are available for VM IO (IOPS consumed by VMs) and
resync IO (IOPS consumed by the backend storage tasks). It all will
happens during dynamic changes and load on the system. The adaptive
resync will guarantee the level of bandwidth to ensure that specific
traffic type will not starve from resources.
根据我的理解，Resync一般发生在数据重构（Rebuild）或者增/删节点时的数据重平衡（Rebalance），Adaptive resync能够保证同步I/O获得至少20%的带宽，而在没有resync流量时则VM I/O可以使用全部100%的带宽。
Faster Destaging（更快的数据下沉）
– VMware has improved the in-memory handling of I/O during destaging.
In fact, during the destage process from the write buffer to the
capacity tier. So the buffer is available for new IOs much quicker.
This optimization provides better performance not only for VM I/O but also for the resynchronization I/O.
上图示例为vSAN全闪存配置，所有写入数据先进入Buffer/Cache SSD分层，然后再destage到容量SSD。VMware应该是改进了内存中的这部分I/O处理，从而提高效率减少拥塞。
扩展阅读《
vForum随笔：全闪存VSAN和Nimble CASL的创新
》
vSAN Replica Consolidation（副本合并）
– It consolidates replicas which are stored across multiple fault
domains (hosts) while maintaining FTT compliance. Before you shut down,
you need to put host to a maintenance mode. You also need to select an
evacuation method (Ensure data accessibility from other hosts, Evacuate
All data to other hosts, or No data evacuation)…
The new improvement:
-
Opens up fault domain as a target for other objects during decommissioning.
-
It also looks for the smallest component to move.
-
As a result, there is fewer data rebuilt and less temporary space usage.
上面的动画有助于理解这部分改进。我们知道在将vSAN集群中的节点撤出之前，需要先进入维护模式，此后面临数据在节点间的移动以保证FTT规则（副本数量不变）。新功能所做的就是减少数据重构量和临时空间占用。
Space efficiency improved via New Object Provisioning
（通过新的对象分配改进空间效率）
– It brings new behaviors with swap file objects for space efficiency. The new swap objects are
thin provisioned（精简配置）
.
-
VM swap now Thin provisioned (previously an advanced parameter had to be set on a per-host basis).
-
A policy will be inherited from VM instead of the Default vSAN storage policy
-
FTT on swap object consistent with applied object policy
如果我没记错的话，
vSAN之前已经允许将swap内存交换磁盘文件设置为thin，在Horizon VDI这样的虚拟机密集型应用中，可以节约不少存储空间
。以前打开这个高级选项还比较麻烦，如今应该成为默认设置了。
Fast Failovers when using Redundant vSAN networks（当使用冗余vSAN网络时快速恢复）
– An Improved resilience of redundant vSAN networks.
New logic has been introduced.
-
Each vmknic is on a different subnet
-
no connectivity between switches
-
There is an immediate failover to other fabric when there is a failure of vmknic in one fabric
Optimized Witness Traffic separation for stretched clusters（针对延伸集群优化的见证节点流量隔离）
– if you’re running an architecture with a stretched cluster
environment, you’ll have a possibility now to benefit from the improved
separation of witness traffic from vSAN data traffic. The new
improvements will prevent data traffic from being routed via vmknic for
witness host traffic when there is a link failure.
It’s configured
by specifying an alternate vmknic (by ESXCLI command only in this
release). This was previously supported already for 2-Node
configurations.
Failure handling improved in case of stretched cluster architecture
（延伸集群架构实例中的故障处理改进）
– you’re in a situation of a site isolation. The inter site link is
down (ISL) but also the link to the third site where the witness is
located, is down as well.
In order to track the component health
properly, in a situation when the connectivity to the Witness site is
back again, vSAN will track the fitness of each site after full
partition event of the Preferred site. The secondary site is maintained
as an Active site until ISL connectivity is restored.
It
prevents false positives
when the ISL is not “up”.
Host Affinity for Next-Gen Applications（针对下一代应用的主机亲和）
– you can now “pin” a VM to a certain node.
The system makes sure that data stays on the same host that VM runs.
针对Hadoop这类有本地I/O优化的应用，vSAN现在
允许限制虚拟机数据存放在本地物理服务器上
，并且以单副本运行（HDFS等自身有多副本保护机制）。
Stretched Cluster Efficient Resync with Proxy Owner（通过代理节点的延伸集群高效重同步）
– this feature’s goal is to reduce inter-site traffic. It minimizes the data which travels across WAN link.
It designs a
proxy owner
on the target site (Site B in our example) which takes advantage of local resync.
-
The partial resync will sync active replica across ISL.
-
It will establish proxy owner to manage local resync.
-
Resync of the replica on site B will occur from a local copy.
在
跨站点（双活）
vSAN部署中，每个站点内部有副本保护，跨站点也是副本保护。在最初的延伸集群中，vSAN似乎对这些副本是“同等对待”的，也就是在恢复B站点的数据时，A站点的2个副本需要分别从Site A复制到Site B完成Resync，这样对数据中心间链路的占用较大。
而在vSAN 6.7版本中，副本数据可以由Site A的
代理Onwer节点跨站点只复制一次
，在Site B的代理Onwer节点
再将其本地Resync复制出多份。这样就节省了WAN链路的数据流量。
Support for Windows Server Failover Clusters (WSFC，故障恢复集群支持)
– now it supports the legacy storage architectures. iSCSI service which
can be activated on the VMware vSAN cluster can present shared storage
to those WSFC servers and the solution is fully supported. It supports
either:
这个是在vSAN 6.5 iSCSI支持基础上的进一步发展。
扩展阅读《
VSAN 6.5详解：传统存储特性附体，未来野心更大
》
vSAN Support Insight
智能支持
– when enabled, the data are anonymized and then uploaded to VMware.
And it is VMware which is able to help when you’re experiencing problem
or performance struggle. The pieces of information (such as names of
hosts, or VMs) are changed before that information is uploaded to VMare.
Health Checks Additions and Improvements（健康检查增加和改进）
– there are some improvements on the health checks for vSAN 6.7 as
well. Those health checks are very useful to troubleshoot or avoid
problems.
-
New host maintenance mode verification ensures proper decommission state
-
New host consistency settings verification for advanced settings
-
Improved vSAN and vMotion Network connectivity checks
-
Improved vSAN health Service Installation check
-
Improved physical disk health check combines multiple checks (software, physical, metadata) into one check
-
Improved HCL check; firmware check now independent from driver check
Enhanced Diagnostic Partition (coredump) Support（增强诊断分区支持）
– for diagnosing issues properly, there is a coredump partition. It is a
special partition for diagnostic purposes. VMware vSAN 6.7 is
automatically extending this partition when needed, depending on the
host requirements. vSAN 6.7 needs larger coredump partition so it is
able to automatically resize the coredump partition
（能够自动调整coredump分区）.
Support for 4Kn Disks
– those are a large disk which has a better price per GB than 512e and
512n disks (supported already by the previous version of VMware vSAN).
New 4kn disks are supported and the vSAN data plane I/O is optimized for
4Kn.
There is a software emulation for 4Kn.
vSAN 6.5已经支持512e扇区格式的HDD机械硬盘，但由于ESXi kernel的原因直到6.7版本才支持原生4kn磁盘。我们知道
512e在未对齐写入时性能会有影响，4Kn则好多了，除非是传统应用发出小于4K的I/O
，这种情况下才可能出现“写放大”。
vSAN的数据路径已经为4Kn优化，这个变化主要影响HDD而与SSD无关。
扩展阅读《
4KB扇区硬盘来了，RAID、VMware兼容不？
》
Improved Security with vSphere FIPS 140-2 validation（数据加密改进安全性）
– VMware VMkernel cryptographic module v1.0 has achieved FIPS 140-2 so
vSAN 6.7 takes benefit of this. It uses this FIPS 140-2 validate d
cryptographic module implemented already in vSphere.
综上所述，vSAN 6.7在之前版本基础上带来了一些重要的改进，可见软件定义存储（SDS）的脚步还会越走越远。
更多相关文章
《
为什么ScaleIO和VSAN不要求三副本？
》
《
存储极客：为什么说VSAN与Dell SC渐行渐近？
》
《
全闪存专享：VSAN 6.2重复数据删除、纠删码浅析
》
注
：本文只代表作者个人观点，与任何组织机构无关，如有错误和不足之处欢迎在留言中批评指正。
进一步交流
技术
，
可以
加我的
QQ/
微信：
490834312
。如果您想在这个公众号上分享自己的技术干货，也欢迎联系我：）
尊重知识，转载时请保留全文，并包括本行及如下二维码。感谢您的阅读和支持！《企业存储技术》微信公众号：
HL_Storage
长按二维码可直接识别关注
历史文章汇总
：
http://chuansong.me/account/huangliang_storage
