vSAN Disk group disks showing 'disks in use' as 0

Products

VMware vSAN

Issue/Introduction

To identify the issue displayed by the vSAN health test result indicating a to memory leak.

Symptoms:

Unable to place a vSAN ESXi host into maintenance mode

Disk group in vSAN shows ‘disks in use’ as 0 in => vCenter > Cluster > Configure > vSAN> Disk management

The Health plugin => vCenter > Cluster > Monitor > vSAN > Health will show different issues every time the health test is run, with a few examples below:

Network - Hosts with connectivity issues
Network - Network latency check
Host configured with different Environment variables

You will see the following error traces in the logs :

# less vsanmgmt.log |grep -i MEMORY
2019-05-21T03:48:53Z VSANMGMTSVC: ERROR vsanperfsvc[Thread-2] [VsanHealthDaemon::doVsanHealthCheck] Get exception in invoking QueryCheckLimits : [Errno 12] Cannot allocate memory
2019-05-21T03:53:53Z VSANMGMTSVC: ERROR vsanperfsvc[Thread-2] [VsanHealthDaemon::doVsanHealthCheck] Get exception in invoking QueryCheckLimits : [Errno 12] Cannot allocate memory
2019-05-21T03:58:53Z VSANMGMTSVC: ERROR vsanperfsvc[Thread-2] [VsanHealthDaemon::doVsanHealthCheck] Get exception in invoking QueryCheckLimits : [Errno 12] Cannot allocate memory
2019-05-21T04:03:53Z VSANMGMTSVC: ERROR vsanperfsvc[Thread-2] [VsanHealthDaemon::doVsanHealthCheck] Get exception in invoking QueryCheckLimits : [Errno 12] Cannot allocate memory
2019-05-21T04:08:53Z VSANMGMTSVC: ERROR vsanperfsvc[Thread-2] [VsanHealthDaemon::doVsanHealthCheck] Get exception in invoking QueryCheckLimits : [Errno 12] Cannot allocate memory
2019-05-21T04:09:11Z VSANMGMTSVC: INFO vsanperfsvc[MainThread] [statsdaemon::_logDaemonMemoryStats] Daemon memory stats: eMin=97.280MB, eMinPeak=103.396MB, rMinPeak=103.396MB

#grep -r "memory stats" var/run/log/vsanmgmt* | grep PRESSURE
No result

Restarting the vSAN mgmt service (vsanmgmt) will fix the issue for a few minutes (10 to 180 minutes) and then issue reappears on the vSAN Health Test page.

On all the hosts in the cluster run:

/etc/init.d/vsanmgmtd restart

You may also verify if the issue is resolved by starting the below services if the "vsanmgmtd" restart does not resolve the issue:
/etc/init.d/vsanvpd restart
/etc/init.d/vpxa restart
/etc/init.d/hostd restart

From vCenter Appliance SSH:

service-control --stop --all
service-control --start –all

Environment

VMware vSAN 6.x

Cause

It’s an issue with Controller Driver "lsi_msgpt3" version (15.00.00.00-1OEM) that is handling the ioctl command from sas3flash due to a known issue.

#/opt/lsi/bin/sas3flash -list
command Hung up

#ps -cPTgjstz | grep sas3flash
101712 101712 sas3flash 101712 18713105 18713105 18713106 U WAIT LOCK 0-23 0.5025 /opt/lsi/bin/sas3flash -list
18714213 18714213 sas3flash 18714213 18713105 18713105 18713106 U WAIT LOCK 0-23 0.5225

#/opt/lsi/bin/sas3flash -list
.
.
Many sas3flash processes waiting.

#ps | grep sas3flash | wc -l
29

A large number of processes running in a WAIT state are causing the vsanmgmt service to crash and the health plugin will display vague results.

The SAS3FLASH plugin initiates the "list "command at regular intervals to the driver, upon not receiving the response the "query" initiated stays in memory. Multiple "queries" remain in memory, finally causing the vsanmgmt service to crash, as can be seen by the ps command mentioned above. The following command can be executed to ensure it is displaying the output without getting hung:

#/opt/lsi/bin/sas3flash -list
Avago Technologies SAS3 Flash Utility
Version 12.00.00.00 (2015.11.19)
Copyright 2008-2015 Avago Technologies. All rights reserved.

Adapter Selected is a Avago SAS: SAS3008(C0)

Controller Number : 0
Controller : SAS3008(C0)
PCI Address : 00:03:00:00
SAS Address : 5001636-0-026d-fe28
NVDATA Version (Default) : 09.00.1a.12
NVDATA Version (Persistent) : 09.00.1a.12
Firmware Product ID : 0x2721 (IR)
Firmware Version : 09.00.00.00
NVDATA Vendor : Quanta
NVDATA Product ID : SAS3008
BIOS Version : 08.21.00.00
UEFI BSD Version : 10.00.00.00
FCODE Version : N/A
Board Name : Quanta Mezz
Board Assembly : 35S2BMA0000
Board Tracer Number : N/A

Finished Processing Commands Successfully.
Exiting SAS3Flash.

Resolution

The issue is fixed in the main branch after the lsi_msgpt3 version 16.00.01.00-7.
The async driver version starting from 17.00.00.00 also contains the fix.
Please make sure the matching firmware version is in place when upgrading the Controller driver.
The fix is expected to be included into vSphere6.5U3 branch.
Reference vSAN HCL : https://www.vmware.com/resources/compatibility/detail.php?deviceCategory=vsanio&productid=39639

Workaround:
Restarting services on the hosts will fix the issue for few minutes only.

Additional Information

Impact/Risks:
No impact on production, but disrupts a few management tasks like placing a host into maintenance mode, failing vSAN health test etc.