vSAN 常見問題集

How does write operation work

Once a write is initiated by the application running inside of the Guest OS, the write is forked to both the write cache on the host which is the ‘owner’ of the storage object, and also to the write cache on the remote host or hosts which contain a replica.

How does cache trigger write back

At regular intervals, the SSD write buffer is de-staged to HDD.

About Number of disk stripes per object

The Number of disk stripes per object setting in the VM Storage Policy means stripe across “at least” this number of magnetic disks per mirror”.

How to perform syncing under abnormal status

components never move

resync->if a server is down

reblanace->if one HDD has over 80% utilisation rate, 20% of data will be moved

absent (default 60mins) will trigger recovery/this is also the windows for offline mantainess

How to test availability

test FT by delete disk group is a correct way

A 4GB VMDK with width set to 2

this VMDK object will be separated into 2 components on 2 Disk Groups.

each of the Disk Group holds 2GB data (4GB/2Disk Groups).

If the Disk Group contains 2 HDDs, whether the data of the component will be separated/distributed on all of these disks (in this case, each Disk holds 1GB data).

Single component cannot spread across multiple HDD's

Are the components composed by blocks? If yes, what's the block size?

1MB/block

When reading data from VSAN, the read request will try to scan write buffer and fetch it form write buffer or not? If not, does it mean when reading the latest written data, the data has to be flushed to HDD and pop up to read cache?

The object logical address is indexed and searchable. The data may be present in memory or on RC or in the write buffer or in the underlying MD. For each incoming read request, I "scan" across the LBA of that read request to find where the most recent data is for each LBA.

To ensure that I return the latest data, I absolutely must scan the write buffer and fetch data from it.

There is no advantage to forcing written data to be flushed to HDD before I refill the cache on a miss.

Nothing pops into the read cache without an explicit read from the guest that causes a read cache miss.

What's the time window for the write buffer to write the data to HDD?

The time window is variable - it is dependent on the size of the SSD and the rate at which write IOs are written to non-overlapping regions, and a few other factors. It can be between tens of minutes to days.

Does the read cache follow move out the old data by FIFO algorithm means it will retire the block which is not being read for the longest time?

Are you asking about the eviction algorithm? If so, then the read cache uses a heavily modified ARC.

RAID 5/6 and Stripe

Also note that if you include Number of disk objects to stripe as a policy setting for the RAID-5/6 objects, each of the individual components that make up the RAID-5 or RAID-6 objects may also be striped across multiple disks.

How is data striped across multiple disks on a host when stripe width is set to 2?

When stripe width is set to 2 first of all there is no guarantee that the data is striped across disks within a host. VSAN has it’s own algorythm to determine where data should be placed and as such it could happen that although you have sufficient disks in all host your data is striped across multiple hosts instead of disks within a host. When data is striped this is done in chunks of 1MB.

Note 1: no guarantee that the new stripe will use a different SSD in different Disk Group

Note 2: no guarantee that the new stripe will use a different HDD in the same Disk Group

Dedupe and Compress

Available on all-flash configuration

With in ONE disk group

Dedupe happens when de-staging data from with cache to data SSD by 4KB block

Multiple VMK

vSAN can have multiple VMK within different VALN, two VMK is used for HA not for load balancing