New 2.6.32 Kernel (pvetest)

[martin]

We just released a new kernel to the pvetest repository.

Release notes:

- pve-kernel-2.6.32-6-pve (2.6.32-48)

• update to vzkernel-2.6.32-042stab039.2.src.rpm

__________________
Best regards,
Martin Maurer

 

[François MASSON]

OCSFS2 support was dropped between pve-kernel-2.6.32-4 and pve-kernel-2.6.32-6, could you please re-anable it as we do use it.
Thanks

 

[dietmar]

OCSFS2 support was dropped between pve-kernel-2.6.32-4 and pve-kernel-2.6.32-6, could you please re-anable it as we do use it.

ask oracle -they discontinued support for standard kernels.

 

[François MASSON]

Argh ! I didn't made the link between the fact that Oracle ceased support and proxmox's kernel. I'll have to stick to 2.6.32-4 til proxmox 2 stable (and complete re-install).
By the way, what clusterfs remain availlable in pve kernel ?

 

[tom]

gfs2. our kernel is based on rhel6x. but we do not use this cluster filesystem on standard installations so do not expect to much help here.

 

[atran]

KSM is not working in this kernel?

"watch cat /sys/kernel/mm/ksm/pages_sharing" shows 0 pages

pve-manager: 1.9-24 (pve-manager/1.9/6542)
running kernel: 2.6.32-6-pve
proxmox-ve-2.6.32: 1.9-48
pve-kernel-2.6.32-4-pve: 2.6.32-33
pve-kernel-2.6.32-6-pve: 2.6.32-48
qemu-server: 1.1-32
pve-firmware: 1.0-14
libpve-storage-perl: 1.0-19
vncterm: 0.9-2
vzctl: 3.0.29-2pve1
vzdump: 1.2-16
vzprocps: 2.0.11-2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.15.0-1
ksm-control-daemon: 1.0-6

 

[tom]

works here.

proxmox-7-106:~# pveversion -v
pve-manager: 1.9-24 (pve-manager/1.9/6542)
running kernel: 2.6.32-6-pve
proxmox-ve-2.6.32: 1.9-48
pve-kernel-2.6.32-4-pve: 2.6.32-33
pve-kernel-2.6.32-6-pve: 2.6.32-48
qemu-server: 1.1-32
pve-firmware: 1.0-14
libpve-storage-perl: 1.0-19
vncterm: 0.9-2
vzctl: 3.0.29-3pve1
vzdump: 1.2-16
vzprocps: 2.0.11-2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.15.0-1
ksm-control-daemon: 1.0-6
proxmox-7-106:~# cat /sys/kernel/mm/ksm/pages_sharing
2488014
proxmox-7-106:~#

 

[atran]

maybe because i have alot of free mem?


host3:~# free
total used free shared buffers cached
Mem: 12285172 6069064 6216108 0 17320 34640
-/+ buffers/cache: 6017104 6268068
Swap: 11534328 0 11534328

 

[tom]

See http://pve.proxmox.com/wiki/KSM

 

[atran]

"...at least 50%..."
OK thanx Tom.

 

[SuSt]

NFS is slightly broken in this kernel. If you mount something (does not matter what) from inside any container, then you can't access to NFS server from HN. And vice versa, if you mount something from HN, then NFS becomes inaccessible from any container with the same message "Address already in use". I'll try to file a bug in OpenVZ's bugzilla.

 

[SuSt]

Please, include this patch into further kernel builds.
http://bugzilla.openvz.org/show_bug.cgi?id=2045

 

[atran]

"...at least 50%..."
OK thanx Tom.

i am now using a lot more mem then 50% but KSM still don't work?!

total used free shared buffers cached
Mem: 12285172 11577344 707828 0 140228 2923552


pve-manager: 1.9-24 (pve-manager/1.9/6542)
running kernel: 2.6.32-6-pve
proxmox-ve-2.6.32: 1.9-48
pve-kernel-2.6.32-4-pve: 2.6.32-33
pve-kernel-2.6.32-6-pve: 2.6.32-48
qemu-server: 1.1-32
pve-firmware: 1.0-14
libpve-storage-perl: 1.0-19
vncterm: 0.9-2
vzctl: 3.0.29-2pve1
vzdump: 1.2-16
vzprocps: 2.0.11-2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.15.0-1
ksm-control-daemon: 1.0-6

host3:~# cat /sys/kernel/mm/ksm/pages_sharing
0

 

[jleg]

i am now using a lot more mem then 50% but KSM still don't work?!

put into /etc/ksmtuned.conf:

LOGFILE=/var/log/ksmtuned
DEBUG=1

restart ksmtuned, and watch /var/log/ksmtuned - after a while it'll tell you why it stopped...

 

[atran]

OK here's the log:

host3:~# tail -f /var/log/ksmtuned
Thu Nov 3 16:37:34 CET 2011: total 12285172
Thu Nov 3 16:37:34 CET 2011: sleep 136
Thu Nov 3 16:37:34 CET 2011: thres 2457034
Thu Nov 3 16:38:34 CET 2011: committed 8043756 free 3745252
Thu Nov 3 16:38:34 CET 2011: 10500790 < 12285172 and free > 2457034, stop ksm
Thu Nov 3 16:39:34 CET 2011: committed 8043780 free 3745380
Thu Nov 3 16:39:34 CET 2011: 10500814 < 12285172 and free > 2457034, stop ksm
Thu Nov 3 16:40:34 CET 2011: committed 8043188 free 3746804
Thu Nov 3 16:40:34 CET 2011: 10500222 < 12285172 and free > 2457034, stop ksm

the server has 12GB RAM.

 

[jleg]

OK here's the log:

host3:~# tail -f /var/log/ksmtuned
Thu Nov 3 16:37:34 CET 2011: total 12285172
Thu Nov 3 16:37:34 CET 2011: sleep 136
Thu Nov 3 16:37:34 CET 2011: thres 2457034
Thu Nov 3 16:38:34 CET 2011: committed 8043756 free 3745252
Thu Nov 3 16:38:34 CET 2011: 10500790 < 12285172 and free > 2457034, stop ksm
Thu Nov 3 16:39:34 CET 2011: committed 8043780 free 3745380
Thu Nov 3 16:39:34 CET 2011: 10500814 < 12285172 and free > 2457034, stop ksm
Thu Nov 3 16:40:34 CET 2011: committed 8043188 free 3746804
Thu Nov 3 16:40:34 CET 2011: 10500222 < 12285172 and free > 2457034, stop ksm

the server has 12GB RAM.

...this means that KSM is explicitly stopped because of this code:

    ...
    free=`free_memory`
    committed=`committed_memory`
    debug committed $committed free $free
    if [ $[committed + thres] -lt $total -a $free -gt $thres ]; then
        KSMCTL stop
        debug "$[committed + thres] < $total and free > $thres, stop ksm"
        return 1
    fi

dunno if this is a bug or a feature of ksmtuned... ;-)

 

[atran]

jleg, thank you for the reply.


can somebody else shine some light on this?
is the if statement correct?

committed and threshold is indeed 10G.

 

[Sascha]

I Just wanted to make sure that one can find this thread using the Keyword OCFS, thus: OCFS OCFS OCFS