DRBD sync problem

[Franck Martinière]

Hello,

I have setup a cluster with two nodes with PROXMOX and DRBD (Primary/Primary).

Virtual machines are localised on DRBD volume and I can be migrated from one node to the other.
For VM, DRBD synchronisation work well.

My problem is when I copy or create a file on the DRBD volume :
If I work on Node 1, I need to enter "sync" command on that node and remount DRBD partition on the second node to see the file from the second node.

My DRBD partition is mounted with "rw,noauto,noatime,sync" options (/etc/fstab).

It seems that on each node drbd volume is buffered in memory ???

If somebody have an explaination or an idea for resolving this matter I'll be very happy.

Thank's in advance,

Franck.

 

[tom]

You mount a partition on both nodes? how? post 'cat /etc/fstab'

 

[macday]

You mount a partition on both nodes? how? post 'cat /etc/fstab'

I did this also ... but only on one node ... to test drbd-performace with pveperf

 

[Franck Martinière]

Yes I mount the DRBD partition on the 2 nodes.

Here is the fstab file :
# <file system> <mount point> <type> <options> <dump> <pass>
proc /proc proc defaults 0 0
# / was on /dev/sda1 during installation
UUID=fa994c39-ff9b-4fec-9630-ef527613075b / ext3 errors=remount-ro 0 1
# swap was on /dev/sda5 during installation
UUID=152614e2-c3cc-4b02-b673-c8f45bae9b22 none swap sw 0 0
/dev/scd0 /media/cdrom0 udf,iso9660 user,noauto 0 0
/dev/sdc1 /media/usb0 auto rw,user,noauto 0 0
/dev/drbd1 /mnt/drbd1 ext4 rw,noauto,noatime,sync 0 0

DRBD partition is the last one.

Franck.

 

[dietmar]

Yes I mount the DRBD partition on the 2 nodes.

Never do that!! This will destroy your data.

 

[Franck Martinière]

Hello Dietmar,

How can I migrate a VM on the second node if DRBD partition is not mounted ?
All my VM are localized on the DRBD.

Best regards,
Franck

 

[tom]

by using LVM on DRBD.

- - - Updated - - -

if you want to mount a partition on several host, you need a cluster file-system, e.g. gfs2. but this is nothing what we recommend here. doing this with ext3 is the end of your data ...

 

[Franck Martinière]

Thank's for your respond.

I have changed my partition from ext4 to GFS2. The original problem is resolved.
When I create a new file I can see it on the other node instantaneously.

But now I can't migrate Virtual machine.
I got this error : "TASK ERROR: command 'clusvcadm -M pvevm:103 -m linux-rd1' failed: exit code 250"

Any suggestion will be appreciated.

Best regards,
Franck.

 

[tom]

your configured HA? post your /etc/pve/cluster.conf

 

[Franck Martinière]

Here is my cluster.conf :

<?xml version="1.0"?>
<cluster config_version="59" name="rdilinux2013">
<cman expected_votes="1" keyfile="/var/lib/pve-cluster/corosync.authkey" two_node="1"/>
<fencedevices>
<fencedevice agent="fence_drac5" ipaddr="172.16.1.55" login="root" name="node1-drac" passwd="rdi0078" secure="1" action="off"/>
<fencedevice agent="fence_drac5" ipaddr="172.16.1.56" login="root" name="node2-drac" passwd="rdi0078" secure="1" action="off"/>
</fencedevices>
<dlm plock_ownership="1" plock_rate_limit="0"/>
<clusternodes>
<clusternode name="linux-rd1" nodeid="1" votes="1">
<fence>
<method name="1">
<device name="node1-drac"/>
</method>
</fence>
</clusternode>
<clusternode name="linux-rd2" nodeid="2" votes="1">
<fence>
<method name="1">
<device name="node2-drac"/>
</method>
</fence>
</clusternode>
</clusternodes>
<rm>
<service autostart="1" exclusive="0" name="VIP" recovery="relocate">
<ip address="172.16.40.250"/>
</service>
<pvevm autostart="0" vmid="101"/>
<pvevm autostart="0" vmid="102"/>
<pvevm autostart="0" vmid="103"/>
<pvevm autostart="0" vmid="104"/>
<pvevm autostart="0" vmid="105"/>
<pvevm autostart="0" vmid="106"/>
</rm>
</cluster>

Best regards,
Franck

 

[Franck Martinière]

After servers restart it appears that online VM migration work.

But when VM is stopped migration is not possible.
We got the same message.
Regards,
Franck

 

[tom]

how do you manage the quorum in your two-node setup?

 

[Franck Martinière]

We don't manage the quorum.
But with Ext4 partition we can migrate VM in stopped mode.

Best regards,
Franck.

 

[tom]

We don't manage the quorum.
But with Ext4 partition we can migrate VM in stopped mode.

Best regards,
Franck.

quorum is essential. so just ignoring it is the reason for a non-working configuration.

if you want HA, follow the recommended setup, otherwise you do not get a stable system.
see http://pve.proxmox.com/wiki/High_Availability_Cluster

two-nodes are never recommended.

 

[Franck Martinière]

Ok Tom our configuration is not perfect.
But for the moment it look like suffisent for our application.

For my question on VM offline migration we have found the respond here : http://188.165.151.221/threads/8995-Online-migration-fails
When VM are managed by HA they can't be migrate offline.

Thank's and best regards
Franck.

 

[tom]

Yes, if your VM is managed by HA (rgmanager), you cannot offline migrate.

See https://bugzilla.proxmox.com/show_bug.cgi?id=297