VMware Cloud Community
stefvanpoucke
Contributor
Contributor

LSOM.ssdWbSize value mismatch

Hi,

This is my first post here since I couldn't find any info about this online or on the forum.

I'm setting up a vSAN Cluster, 2 hosts + witness appliance.

The witness appliance (10.0.99.120) has a property value mismatch for LSOM.ssdWbSize.

Under info it refers to this vmware kb : https://kb.vmware.com/s/article/2107713 Here it is mentioned that the setting should be in Configure -> System -> Advanced system settings'

However I can't find it there. Any advice or workaround would be welcome.

pastedImage_0.png

0 Kudos
7 Replies
lucasbernadsky
Hot Shot
Hot Shot

Hi there! I've never heard about that alert.

It seems to be a cache / capacity ratio problem. vSAN require the cache disk to have -at least- 10% size of the capacity tier. For example, for a capacity tier of 2TB, the cache disk must be 200GB or bigger.

Wb seems to be Write Buffer.

I found this chinese site explaining the advanced settings, you can translate it to english using the browser: https://www.halfcoffee.com/2016/09/12/vsanlsom/

Also this post may be useful Write Buffer Sizing in vSAN when Using the Very Latest Hardware

If you run vsan debug advcfg list in an SSH session, you will see that option listed.

For example:

vsan debug advcfg list --option='LSOM.ssdWbSize'

Regards

0 Kudos
TheBobkin
Champion
Champion

Hello stefvanpoucke​,

Welcome to Communities.

Is the Witness running on a different build of ESXi than the data-nodes are?

I ask as (other than manual advanced configuration changes) this is the most common source of such alerts in vSAN Health - this is as a result of ESXi/vSAN having different parameters set for certain things in different build versions, not all of these are customer-facing and thus not all configurable via the UI (as some of these really shouldn't be messed with).

If it is running on a different build then update the nodes to the matching build or if this is not possible then redeploy the Witness on the same ESXi build the hosts have installed.

Bob

0 Kudos
stefvanpoucke
Contributor
Contributor

Hi lucas,

On the datanode's I have 4x 1.8TB disks and a 900 GB SSD so that would be around 12.5%

I don't really know how this works with the witness appliance, with the deployment I just configured one for 500VM's.

pastedImage_0.png

I SSH'd into the machine's but the command was unknown.

TheBobkin​ I checked the builds and you were right

the 2 data node's and the server running the witness appliance are V6.7 Build 15160138

The witness appliance itself is V6.7 Build 8169922.

I've reinstalled the witness appliance with the correct build.

but... the issue still persists

0 Kudos
TheBobkin
Champion
Champion

Hello stefvanpoucke,

So I had a quick dig into this from internal documentation and it would appear that LSOM.ssdWbSize=90 is the correct setting for a Witness with default settings, so this health alert is not a cause for concern.

However, the check for this in Health is intended to filter out this check on the Witness (as it is an intentional difference and thus we don't want it triggering a warning) - what build of vCenter is in use here?

If the hosts are on 6.7 U3 then so should vCenter:

VMware Product Interoperability Matrices

Bob

0 Kudos
stefvanpoucke
Contributor
Contributor

Hi Bob,

the vCenter version matches with the rest.

According to the matrix I could also use V7.0? maybe I should do that..

If the error persists should I just silence it and try to continue with the vSAN config?

0 Kudos
TheBobkin
Champion
Champion

Hello stefvanpoucke

If you haven't restarted services and/or rebooted the vCenter since updating the Witness I would try this first if even just to rule this out.

I don't think updating to 7.0 will fix this as from vSAN Health perspective this has mostly the same functionality/code as 6.7 U3 (7.0 GA anyway).

I wouldn't advise silencing this Health check as it is responsible for checking a lot of other configuration settings and can be quite useful/informative.

As I said above - that setting for the Witness is correct and should pose no functional/operational issues, the issue is with vSAN Health not seeing that it is a Witness and should be ignoring the disparity.

"If the error persists should I just silence it and try to continue with the vSAN config?"

Has the cluster been fully configured e.g. all nodes have vSAN networking, Disk-Groups and the Witness (10.0.99.120) added? If not, please complete the configuration.

If this is a production cluster and/or you have a support contract with VMware GSS I would advise opening a Support Request with us so that we can figure out why vSAN Health is triggering this.

Bob

0 Kudos
elies_jebri
Contributor
Contributor

Modify the default ratio of Read Cache and Write Buffer in VSAN to 7:3

VSAN's default Read Cache and Write Buffer are provided by the same flash memory device, and many people have heard that this ratio is 7:3, but how many people know how this ratio comes from?

First of all, the biggest reason for the existence of the 7:3 ratio is because the occurrence rate of Read IOPS in tiered storage is much higher than that of Write. Therefore, VMware's official default setting will only guarantee this 7:3. ratio. This parameter is actually controlled by a parameter "ssdWbSize" located in the VMkernel. The default value of this parameter is 30, and the calculation unit is a percentage. By controlling this percentage parameter, the ratio of Write Buffer can be adjusted;

Then, if you need to adjust the read-write ratio, you can modify the parameter "ssdWbSize". For example, you can modify it to 20 and adjust the read-write ratio to 8:2. The steps are as follows:

1. Execute the vsish (Vmkernel System Information Shell) command in the ESXi Shell of the ESXi host to view the current ratio:
#vsish
›cat /config/LSOM/intOpts/ssdWbSize
Vmkernel Config Option {
Default value:30
Min value:10
Max value:90
Current value:90
hidden config option:1
Description:SSD Write Buffer size in percentage
Host specific config option:0
Option update requires reboot:0
Option update requires maintenance mode:0
}
Remark:
• The modification limit of this parameter can only be 9:1, and it is impossible to give 100% to anyone;
• The parameter "ssdWbSize" is hidden by default;

2. Next, execute the following command to modify the Write Buffer ratio:
#esxcfg-advcfg -s 10 /LSOM/ssdWbSize
OR, you can also use the following command:
#esxcli system settings advanced set -o /LSOM/ssdWbSize -i 30

 

0 Kudos