[pve-devel] memory consumption

Tue Aug 18 20:28:11 CEST 2015

>>To be more precise some values like network ip's etc still are shown and seem to work but not the cache setting and also backbill settings

The cache setting is client side setting, monitors don't known nothing about it.

----- Mail original -----
De: "VELARTIS Philipp Dürhammer" <p.duerhammer at velartis.at>
À: "aderumier" <aderumier at odiso.com>
Cc: "Eneko Lacunza" <elacunza at binovo.es>, "pve-devel" <pve-devel at pve.proxmox.com>
Envoyé: Mardi 18 Août 2015 20:20:30
Objet: AW: [pve-devel] memory consumption

OK, 

i was testing around a bit and my conclution is: 
because of setting the cache size too big the vms used a lot more memory then expected. 
1) because oft he bug described by alexandre changing the cache setting didn't change anything 
2) my ceph.conf does not show in " ceph -n mon.2 --show-config " 
3) but still seems to make effect on the cheph osd nodes 
4) but not on ceph client nodes! 

So on ceph osd node the ceph rbd cache size changes seem to make changes - but not on other nodes where i just use the ceph client to mount the rbd 

So there are two thing that i dont understand: 
>>> why is "ceph -n mon.2 --show-config" not showing the ceph.config values - but still makes effect (and only!) on nodes with ceph osd's and mons but not on client nodes? 

To be more precise some values like network ip's etc still are shown and seem to work but not the cache setting and also backbill settings 

-----Ursprüngliche Nachricht----- 
Von: Alexandre DERUMIER [mailto:aderumier at odiso.com] 
Gesendet: Dienstag, 18. August 2015 20:08 
An: VELARTIS Philipp Dürhammer 
Cc: Eneko Lacunza; pve-devel 
Betreff: Re: [pve-devel] memory consumption 

>>if cache=none is set this should overwrite the ceph config from rbd 
>>cache = true=> false but still this machine is consuming 20G instead 
>>of 16G 

Be carefull, they are a bug in current qemu librbd, setting "rbd_cache=true|false" in ceph.conf, always override qemu cache setting. 

This is fixed in qemu 2.4 (for proxmox4). 

But for the moment, the best way is to remove "rbd_cache" from ceph.conf. 

----- Mail original ----- 
De: "VELARTIS Philipp Dürhammer" <p.duerhammer at velartis.at> 
À: "VELARTIS Philipp Dürhammer" <p.duerhammer at velartis.at>, "Eneko Lacunza" <elacunza at binovo.es>, "pve-devel" <pve-devel at pve.proxmox.com> 
Envoyé: Mardi 18 Août 2015 16:34:53 
Objet: Re: [pve-devel] memory consumption 

Hi, 

debugging and reading more 

if cache=none is set this should overwrite the ceph config from rbd cache = true=> false 

but still this machine is consuming 20G instead of 16G 

Von: pve-devel [mailto:pve-devel-bounces at pve.proxmox.com] Im Auftrag von VELARTIS Philipp Dürhammer 
Gesendet: Dienstag, 18. August 2015 16:01 
An: 'Eneko Lacunza'; pve-devel at pve.proxmox.com 
Betreff: Re: [pve-devel] memory consumption 

Hi Eneko! 

I use Proxmox Ceph Server. 

I found out that my config for rbd cache size was too big which could explain me the problem. 

But when i make: 

ceph -n mon.2 --show-config 

it shows me the standard config size from ceph. And also that rbd cache = false. Very strange. 

Is this not the way to get the running conf? 

My client section in the /etc/pve/ceph.conf looks like this: 

[client] rbd cache = true 

rbd cache size = 268435456 

rbd cache max dirty = 33554432 

rbd cache max dirty age = 3 

but seems to be ignored? 

I am using cache=writeback and cache=none. Same results 

Von: pve-devel [ mailto:pve-devel-bounces at pve.proxmox.com ] Im Auftrag von Eneko Lacunza 
Gesendet: Dienstag, 18. August 2015 14:36 
An: pve-devel at pve.proxmox.com 
Betreff: Re: [pve-devel] memory consumption 

Hi Philipp, 

What Ceph packages? 

We have 3 small clusters using creph/rbd and I'm not seeing this. Biggest VM is 6GB of RAM though, biggest disk is 200GB and a VM has a total of 425GB in 5 disks. 

Are you using Proxmox Ceph Server? 
What cache setting in VM harddisk config? 

Cheers 
Eneko 

El 18/08/15 a las 14:22, VELARTIS Philipp Dürhammer escribió: 

Hi, 

some of my virtual machines on different servers are consuming a lot more ram then they should! 

(up to twice) which is a horror if the machines have about 15-20G but consume 30-40GB 

I was checking: 

) all machines have the same pve packages installed 

) it only happens on machines with ceph osd installed 

) which also are using rbd client for virtual hdds 

It does not happen if they are using nfs backend for virtual hdd. Or other machines which are just using the rbd client but doent have osd’s… 

Has somebody experienced something like this also? 

Any ideas how to debug further? 

Machines config: 

proxmox-ve-2.6.32: 3.3-147 (running kernel: 3.10.0-7-pve) 

pve-manager: 3.4-9 (running version: 3.4-9/4b51d87a) 

pve-kernel-3.10.0-7-pve: 3.10.0-27 

pve-kernel-2.6.32-32-pve: 2.6.32-136 

pve-kernel-2.6.32-30-pve: 2.6.32-130 

pve-kernel-2.6.32-37-pve: 2.6.32-150 

pve-kernel-2.6.32-29-pve: 2.6.32-126 

pve-kernel-3.10.0-3-pve: 3.10.0-11 

lvm2: 2.02.98-pve4 

clvm: 2.02.98-pve4 

corosync-pve: 1.4.7-1 

openais-pve: 1.1.4-3 

libqb0: 0.11.1-2 

redhat-cluster-pve: 3.2.0-2 

resource-agents-pve: 3.9.2-4 

fence-agents-pve: 4.0.10-3 

pve-cluster: 3.0-18 

qemu-server: 3.4-6 

pve-firmware: 1.1-4 

libpve-common-perl: 3.0-24 

libpve-access-control: 3.0-16 

libpve-storage-perl: 3.0-33 

pve-libspice-server1: 0.12.4-3 

vncterm: 1.1-8 

vzctl: 4.0-1pve6 

vzprocps: 2.0.11-2 

vzquota: 3.1-2 

pve-qemu-kvm: 2.2-11 

ksm-control-daemon: 1.1-1 

glusterfs-client: 3.5.2-1 

_______________________________________________ 
pve-devel mailing list 
pve-devel at pve.proxmox.com 
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 

-- 
Zuzendari Teknikoa / Director Técnico 
Binovo IT Human Project, S.L. 
Telf. 943575997 
943493611 
Astigarraga bidea 2, planta 6 dcha., ofi. 3-2; 20180 Oiartzun (Gipuzkoa) www.binovo.es 

_______________________________________________ 
pve-devel mailing list 
pve-devel at pve.proxmox.com 
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 