VCSA 6.5 - memory allocation inconsistencies

Czernobog · ‎12-12-2018

Following some performance issues I have checked the memory allocation on some of my VCSA's. The active memory usage seems to be within limits, however there are inconsistencies in the output of the different monitoring tools.

We have assigned 48GB physical memory to the appliance. The appliance is an upgraded 6.0 VCSA, a 'large' deployment and we are within limits of the sizing recommendations (<1000 hosts, <10000 vms), with external PSCs.

Here's the command output:

This one looks good:

root@vcsa [ ~ ]# free -m

total used free shared buff/cache available

Mem: 48294 12085 9336 2279 26872 33606

Swap: 52215 0 52215

edit: Here's the output from top:

PID	USER	PR	NI	VIRT	RES	%CPU	%MEM	TIME+	S	COMMAND
3659	vpxd	20	0	8642.3m	2.309g	13.8	4.9	1128:07	S	vpxd
54010	vpostgr+	20	0	2442.6m	2.210g	0.7	4.7	8:03.53	S	postgres
1664	vpostgr+	20	0	2377.3m	2.162g	0.0	4.6	13:15.61	S	postgres
1665	vpostgr+	20	0	2376.4m	1.956g	0.0	4.1	0:36.64	S	postgres
54011	vpostgr+	20	0	2442.6m	1.799g	0.0	3.8	0:45.09	S	postgres
1637	vsphere+	20	0	11.481g	1.446g	0.7	3.1	79:39.94	S	vsphere-client.
1629	vsphere+	20	0	11.411g	1.197g	1.3	2.5	49:37.24	S	vsphere-ui.laun
2585	root	20	0	10.149g	913.2m	0.7	1.9	57:05.21	S	vmware-vpxd-svc
4455	content+	20	0	9623.1m	848.8m	0.0	1.8	24:08.65	S	vmware-content-
4450	root	20	0	12.064g	622.3m	1.3	1.3	63:36.82	S	vmware-sps.laun
2584	vapiEnd+	20	0	8237.6m	617.2m	0.7	1.3	43:36.80	S	vmware-vapi-end
5361	updatem+	20	0	17.146g	589.1m	0.0	1.2	10:52.27	S	java
58548	vpostgr+	20	0	2381.5m	503.0m	0.0	1.0	0:02.56	S	postgres
1608	root	20	0	4382.2m	442.9m	0.0	0.9	12:27.50	S	vmware-cm.launc
4467	perfcha+	20	0	8273.7m	426.4m	0.0	0.9	10:23.85	S	vmware-perfchar

The dynamic allocation seems to work too:

root@vcsa [ ~ ]# cloudvm-ram-size -l
vmcad                =      224
vmafdd               =       71
vmware-rbd-watchdog  =      100
applmgmt             =      376
vmware-vsan-health   =      117
vmware-vsm           =      273
vmware-sps           =     3476
vmware-stsd          =     1678
vmware-vpostgres     =     4334
vmware-eam           =      323
vmware-sts-idmd      =      483
vmware-mbcs          =      128
vcha                 =       46
vmware-vmon          =        5
vmware-statsmonitor  =       10
vmware-perfcharts    =     1704
vsphere-client       =     2510
vmonapi              =       15
vmware-cm            =      574
vmware-rhttpproxy    =      152
vmware-imagebuilder  =       50
vmware-sca           =      128
vmware-vpxd          =    19330
vsphere-ui           =     2510
vmware-vapi-endpoint =      256
vmware-content-library =      887
vmdnsd               =       72
vmware-updatemgr     =     1046
vmware-vpxd-svcs     =     1459
OS                   =     5925
vmware-netdumper     =       20
TOTAL(MB)            =    48282

Now, the output from cloudvm-ram-size -S is what is weird, because it seems we would need to increate the physical RAM to over 70GB, to meet the service needs:

Service-Name AllocatedMB MaxMB CurrentMB Curr-RSS Cache MapFiles MemoryLimit

LinuxKernel -1 35314 35592 0 652 5 8796093022207

vmware-vpostgres 4334 13854 13842 178 13663 2283 8796093022207

vpxd -1 4005 3399 2287 1111 68 8796093022207

vpxd-svcs -1 972 958 915 42 2 8796093022207

vsphere-client 2510 2428 2112 1536 576 0 8796093022207

vsphere-ui 2510 2093 1828 1286 541 0 8796093022207

TOTAL(RAM=48294MB) 9821 70626 70530 11642 24413 2548 8796093022207

The 'AllocatedMB' value is inconsistent, with what cloudvm-ram-size -l delivers. Why is that?

Schloud I increase the memory allocation by hand with cloudvm-ram-size -C ?

added top output