Start a Conversation

Unsolved

This post is more than 5 years old

J

8223

July 18th, 2017 10:00

Changing datapage size from 2MB to 512k

Hi. I am considering, but not even leaning to yet, changing my datapage size from 2MB to 512k. The reason being is with my Unmap issues I am having, it is possible that the Unmap on VMFS6 will not work unless your array is at a 512k datapage size. That has not been confirmed yet. So please do not take this as truth yet. But I am going on the assumption that may be in issue, so I am starting to think what it will take, and the ramifications of using 512k.

So from the documentation:

By default, data is migrated between tiers and RAID levels in 2 MB blocks. Data can be moved in smaller or larger blocks to meet specific application requirements. These blocks are referred to as datapages.
• 2 MB: Default datapage size, this selection is appropriate for most applications.
• 512 KB: Appropriate for applications with high performance needs, or in environments in which Replays are taken frequently under heavy IO. Selecting this size increases overhead and reduces the maximum available space in the Storage Type.
NOTE: By default, the datapage size is 512 KB when SSDs are present. You can change the page size to 2 MB when you create a disk folder.
• 4 MB: Appropriate for systems that use a large amount of disk space with infrequent Replays.

We have several compellents, each of course with an SSD tier. A couple with optimized flash tier.

So what is strange is that the documentation says "By default, the datapage size is 512 KB when SSDs are present. You can change the page size to 2 MB when you create a disk folder" However our installers always set 2Mb storage type. Even with SSD.

We also take 15min replay, which expire every hour. So that seems to say a smaller block size would help.

Anyway I am trying to decide what are the pros/cons of changing this. And I realize it is probably a ton of work in terms of unassiging disks and creating new disk folder. etc..

So if any comments on this page size change please chime in. Thank you.

July 26th, 2017 18:00

The issues is that the Automatic UNAMP for esx is not working even when the block size of the VMFS6 volume is 1MB and the granularity is also set to 1MB.

According to VMware this feature is not supported at this time on Arrays such as the Compellent arrays.

This is on the release notes for ESX 6.5

pubs.vmware.com/.../vsphere-esxi-vcenter-server-65-release-notes.html

Automatic UNMAP is not supported on arrays with UNMAP granularity
greater than 1MB.

And it is further explained in the following article.

cormachogan.com/.../

17 Posts

January 18th, 2021 11:00

Appreciate this is old thread.

But unluckily we are in this situation at the moment. We have SC Series 5200 active-active.

One Storage controller is 512KB and other 2 MB (no one knows) why the other storage controller is configured in this manner. We also observe no Unmap is happening and we are loosing 10 TB plus space on this storage controller. We are also running 6.5 U3. In short scenario is completely matching.

Question: Is this fixed in 6.7? I did not saw any release notes on it

Question: Is it possible fix this problem without vacating Disks from this folder?

Highly appreciate any inputs or directions

Moderator

 • 

7.6K Posts

January 18th, 2021 12:00

Hello PMZara,

In most cases when I have seen different data pages it is due to different data types.  Here is the link to & if you look on page 40 it states the following:

Datapage Size

By default, data is migrated between tiers and RAID levels in 2 MB blocks. Data can be moved in smaller or larger blocks to meet specific application requirements. These blocks are referred to as datapages.

•2 MB: Default datapage size, this selection is appropriate for most applications.

•512 KB: Appropriate for applications with high performance needs, or in environments in which snapshots are taken frequently under heavy IO. Selecting this size increases overhead and reduces the maximum available space in the Storage Type. Flash Optimized storage types use 512 KB by default.

•4 MB: Appropriate for systems that use a large amount of disk space with infrequent snapshots.

 

CAUTION: Before changing the datapage setting, contact Dell Technical Support to discuss the impact on performance and for advice about how to ensure that system resources remain balanced.

https://dell.to/3io1lyx

 

What is your current SCOS version?

17 Posts

January 21st, 2021 08:00

Sam,

Thanks a lot for the information.

We did opened the case on it.

No Events found!

Top