Memory size reported by MAAS less than actual memory size by 1GB for commissioned VM

[Larry Michel]

I deployed a charm multiple times with memory constraint set to 16GB. Then, it failed with MAAS returning error that no server was available. I checked in MAAS and now server is showing 15GB.

I have recommissioned another VM and it's also showing 15GB.

Looking at a previous juju status output, I see that it's reporting less then 16384M:

  "1":
    agent-state: started
    agent-version: 1.25.3
    dns-name: euclidvm01.oilstaging
    instance-id: /MAAS/api/1.0/nodes/node-ac14c5a4-c154-11e5-98d5-00163e35cf93/
    series: trusty
    hardware: arch=amd64 cpu-cores=4 mem=16049M tags=jujutest

[Larry Michel (lmic)]

Looking at my original bundle, it was set to 12GB so I can't really say it was ever set to 16GB in maas. I recall seeing 16GB but I don't have a way to confirm I observed visually.

[Andres Rodriguez (andreserl)]

Hi Larry,

Do the machines actually have 16GB of RAM?

16049/1024 = 15.67851

Maybe, we are rounding to the exact value 15GB, and not to 16GB ?

[Larry Michel (lmic)]

Hi Andres, lshw also shows 15GB and /proc/meminfo shows a value closer to 16GB so maas must be going by lshw?

ubuntu@euclidvm02:/var/log$ sudo lshw -class memory
  *-firmware
       description: BIOS
       vendor: SeaBIOS
       physical id: 0
       version: 1.7.5-20150306_163512-brownie
       date: 04/01/2014
       size: 96KiB
  *-memory
       description: System Memory
       physical id: 1000
       size: 15GiB
     *-bank
          description: DIMM RAM [empty]
          vendor: QEMU
          physical id: 0
          slot: DIMM 0

My 8G are reporting 8G so this is strange that this is happening only on the 16G VMs. I guess we can mark it as invalid for maas since that would be what maas goes by.

Then,

ubuntu@euclidvm02:/var/log$ cat /proc/meminfo
MemTotal: 16434168 kB
MemFree: 11159676 kB
Buffers: 111044 kB
Cached: 2711004 kB
SwapCached: 0 kB
Active: 3647908 kB
Inactive: 1096152 kB
Active(anon): 1931112 kB
Inactive(anon): 8252 kB
Active(file): 1716796 kB
Inactive(file): 1087900 kB
Unevictable: 3704 kB
Mlocked: 3704 kB
SwapTotal: 4194300 kB
SwapFree: 4194300 kB
Dirty: 3344 kB
Writeback: 0 kB
AnonPages: 1925752 kB
Mapped: 66020 kB
Shmem: 14896 kB
Slab: 354900 kB
SReclaimable: 230392 kB
SUnreclaim: 124508 kB
KernelStack: 24728 kB
PageTables: 31924 kB
NFS_Unstable: 0 kB
Bounce: 0 kB
WritebackTmp: 0 kB
CommitLimit: 12411384 kB
Committed_AS: 62323764 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 38152 kB
VmallocChunk: 34359693608 kB
HardwareCorrupted: 0 kB
AnonHugePages: 1421312 kB
HugePages_Total: 0
HugePages_Free: 0
HugePages_Rsvd: 0
HugePages_Surp: 0
Hugepagesize: 2048 kB
DirectMap4k: 63344 kB
DirectMap2M: 16713728 kB

[Lee Trager (ltrager)]

This is due to lshw changing out it outputs memory between Trusty and Xenial.

On Trusty I get
$ sudo lshw -class memory
  *-firmware
       description: BIOS
       vendor: SeaBIOS
       physical id: 0
       version: Ubuntu-1.8.2-1ubuntu1
       date: 04/01/2014
       size: 96KiB
  *-memory
       description: System Memory
       physical id: 1000
       size: 2GiB
       capacity: 2GiB
     *-bank
          description: DIMM RAM
          vendor: QEMU
          physical id: 0
          slot: DIMM 0
          size: 2GiB

On Xenial I get
$ sudo lshw -class memory
  *-memory
       description: System memory
       physical id: 0
       size: 2000MiB

[Lee Trager (ltrager)]

lshw has a bug on this https://bugs.launchpad.net/maas/+bug/1039701