All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Why are I'm loosing RAM on one server (down 30 MB last minute, 400 MB last 4 days)
Have a strange issue at one of my VPS servers. It's a KVM cloud server that I have in the US. Have not used it for almost 6 months, start using it some days ago.
The server has 2 GB of RAM.
After a reboot 4 days ago, all as normal. Virtualmin tells me that I have 1.85 GB of real RAM on the server. Thats normal I think, Virtualmin say the same on my iwStack 2 GB KVM servers also.
But each day I have lost RAM on the server. Today I get this:
Real memory 1.47 GB total, 295.78 MB used
Mem: 1610976k total, 1187776k used, 423200k free, 156144k buffers Swap: 2326520k total, 0k used, 2326520k free, 797716k cached
It's not just Virtualmin as you can see from SSH output also. It's down 400 MB. Just the last minute after I started writing this, I have lost 30 MB more.
Why?
The server is running CentOS 6.5 64bit with Virtualmin 4.04. It's a KVM cloud.
Comments
erm... yeah that's not right haha, unless they have some sort of dynamic allocation for load balancing.
what is the output of 'cat /proc/meminfo' ?
Ant.
MemTotal: 1439216 kB MemFree: 250688 kB Buffers: 156216 kB Cached: 798216 kB SwapCached: 0 kB Active: 474440 kB Inactive: 586516 kB Active(anon): 22820 kB Inactive(anon): 85288 kB Active(file): 451620 kB Inactive(file): 501228 kB Unevictable: 0 kB Mlocked: 0 kB SwapTotal: 2326520 kB SwapFree: 2326520 kB Dirty: 4 kB Writeback: 0 kB AnonPages: 106480 kB Mapped: 19608 kB Shmem: 1584 kB Slab: 76324 kB SReclaimable: 47284 kB SUnreclaim: 29040 kB KernelStack: 1840 kB PageTables: 10676 kB NFS_Unstable: 0 kB Bounce: 0 kB WritebackTmp: 0 kB CommitLimit: 3046128 kB Committed_AS: 379296 kB VmallocTotal: 34359738367 kB VmallocUsed: 24108 kB VmallocChunk: 34359702420 kB HardwareCorrupted: 0 kB AnonHugePages: 14336 kB HugePages_Total: 0 HugePages_Free: 0 HugePages_Rsvd: 0 HugePages_Surp: 0 Hugepagesize: 2048 kB DirectMap4k: 8180 kB DirectMap2M: 2351104 kBThe RAM is down more:
Mem: 1442156k total, 1189204k used, 252952k free, 156220k buffersReboot again and see with how much it starts?
Will do..
Yeh, that is not right at all.
Probably worth a ticket to the provider.
erm... what? "DirectMap4k: 8180 kB" LOL
get a ticket in, I don't know who the host is but it seems they only guarantee your server around 8mb of ram hehe.
Mem: 1506412k total, 356844k used, 1149568k free, 25192k buffersReal memory 1.44 GB total, 245.46 MB usedStrange, it was 1.85 GB 4 days ago.
WTF...not good.
It's a promotion server, I'm only paying $1/mo for it (after a big promotion) and the diskspeed is insane (over 1 GB/s). But something wrong with it now I can see.
Will try to open a ticket.
to be fair it could just be some funky reporting if they are using an embeded kernel or something but worth asking.
You could also try using all your ram to see when it starts swapping like crazy.
Is this what Xen ballooning looks like?
Memory ballooning.
If you want to be annoying, reinstall the kernel and remove the virtio balooning. That shows him
Lol... but serious
Real memory 1.34 GB total, 269.95 MB usedHow? I can't reinstall OS on that server, have to open a ticket. (I know..thats why I'm only paying $1/mo and are happy with it
)
Yep, but this is KVM so I am just taking my best guesses
this is why I always disagree with people who say you can oversell ram in Xen because it becomes obvious and apparent with balooning you cant allocate 32 VM's 2GB on a 32GB node because the amount of Ram you would need to reserve for balooning would prevent 50% of them from booting to begin with and what ever did boot would show a decreased Ram amount. unlike KVM which will just start swapping.
Well, $1/mo can't buy you 2GB of real RAM so i guess you'll have to accept the ballooned RAM for that price.
True. The plan was a $50/mo plan. They dropped the price to $25/mo (for normal paying customers). Now they do not have the plan anymore.
And how do we know this is KVM? @myhken can you please "lspci" for us?
-bash: lspci: command not found
May I suggest you check that your memory modules are properly clamped. Because if your memory is "loose," you'll soon suffer memory corruption.
As to you thinking you're "losing" memory, top is pretty terrible at giving the details on cached and shared memory. Especially when you take into account drivers in the kernel, kernel log messages, and memory allocated for the graphics driver.
Unless you're noticing poor performance, don't worry about it.
@myhken what control panel are they providing?
could you run:
uname -a
df -h
cat /proc/fstab
Hmmm...maybe not a KVM? Can't find any info about it on their page just now. Maybe I just think it's a KVM
Found this on my control panel
Memory 2304 MBlol.No, that's pretty typical for a KVM. An OpenVZ or container based VPS wouldn't have the kernel and driver overhead (and the related complexity).
Some homemade stuff..
[root@xx ~]# uname -a Linux xxxx 2.6.32-431.3.1.el6.x86_64 #1 SMP Fri Jan 3 21:39:27 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux[root@xx ~]# df -h Filesystem Size Used Avail Use% Mounted on /dev/mapper/vg_xx-lv_root 27G 3.3G 23G 13% / tmpfs 1.1G 0 1.1G 0% /dev/shm /dev/vda1 485M 97M 363M 21% /boot[root@xx ~]# cat /proc/fstab cat: /proc/fstab: No such file or directoryNevermind, i was confused.
It's /proc/mounts btw
Try run "lscpu" to check the virtualization type.
Cloudive kvm vps actually use memory ballooning too. I know because i ask the provider itself.
Here is the CP:
sorry I meant /etc/fstab
But never mind, its obviously KVM.
Only the host will be able to tell you what the situation is, they obviously have a custom implementation ballooning does not exist on KVM in the exact same way as Xen, it is more automatic, essentially it reclaims from guest servers when it needs to/ gets desperate.
The host probably just needs to move a few VPS's round to rebalance things, it is as a result of over allocation on KVM.
I would agree with someone of the other comments that said not to worry about it but given how dynamic this issue is personally I think you do need to worry about it.