This is expected because of the way NIOS handles memory. The CLI command does not show the actual memory utilization, what you see in GUI is accurate. See below for more details
----------------------------------
The memory utilization dashboard widget is different from that which is shown when executing the "show memory" command on the CLI of NIOS appliance. The calculations done for each respective value differ. Furthermore, with the move to the usage of HugePages in 8.1.6, the derivation of the dashboard memory utilization widget value is less obvious.
**Note the memory utilization dashboard widget and the value of the ibSystemMonitorMemUsage (OID: 1.3.6.1.4.1.7779.3.1.1.2.1.8.2.1) are the same value and calculated in the same manner.
Prior to NIOS 8.1.6, the calculation of free memory is:
freeMem = MemFree + cachedMemory + bufferMemory + slabMemory
From NIOS 8.1.6 and beyone the calculation of free memory is:
freeMem = MemFree + cachedMemory + bufferMemory + slabMemory + (hugePages * hugePageSize)
In all cases, the memory utilization value is calculated as such:
MemUsed = ((TotalMem - freeMem)/TotalMem)*100
The show memory command displays the following information:
Infoblox > show memory
total used free buffers cached
Mem: 12176268 10947772 1228496 284024 3538472
Swap: 12582908 0 12582908
Total: 24759176 10947772 13811404
The slab and HugePages information is missing from the ouput just above. Therefore, it is not possible to perform the calculation to match the Dashboard Memory Utilization value using this output.
Before 8.1.6, the only value missing was slab, which tends to be a small value. One can get a close approximation without having the slab memory number for pre-8.1.6 systems.
After 8.1.6, HugePages are used and are a significant value. Additionally, the cache memory value decreases in part because of the usage of HugePages.
**Note: In the Linux kernel, cached memory includes shared memory. What was before considered to be shared memory is now part of the HugePages allocation. Decreasing shared memory, decreases cached memory and would therefore, make the calculation using the "show memory" output inaccurate by a larger margin.
