VZ Container Failover and Migration with ISCSI

[thesubmitter]

I have 4 ProxMox Servers

1. Each of them has a different LUN of 100GB mounted to /data/containers
2. Each of them hosts different containers
3. In the event of migration, I use the regular method and there is an ISCSI --> HOST --> ISCSI Data transfer (which is not optimal but works)

4. In the event of the failure (total explosion) of one of my HW Nodes, I just relocate the corresponding config files from /etc/pve/nodes/FAILEDHOST/openvz/1xx.conf TO /etc/pve/nodes/LIVEHOST/openvz/1xx.conf and start up the containers.

5. Optionally, if I want to restart all the down containers on the SAME BACKUP HW NODE then I can create a NEW LOCAL DIRECTORY TYPE STORAGE DEVICE---> corresponding to the same ISCSI LUN on a new mountpoint say /data/containers2
and then also edit the *.conf files switching out the storage names.

Is anything missing here? I have tested it and it seems to work....

 

[dietmar]

You mount shared iSCSI with ext3 on several nodes? If so, that is a very bad idea.

 

[thesubmitter]

No, only in the case of failure i might create a new storage entry on one of the alive servers and mount the failed servers LUN there and modify the container conf files to point to that new container.

 

[thesubmitter]

I left out the part where I copy the data from the failed server LUN to another LUN mounted by a living server. I added it below

I have 4 ProxMox Servers

1. Each of them has a different LUN of 100GB mounted to /data/containers
2. Each of them hosts different containers
3. In the event of migration, I use the regular method and there is an ISCSI --> HOST --> ISCSI Data transfer (which is not optimal but works)

4. In the event of the failure (total explosion) of one of my HW Nodes, I just relocate the corresponding config files from /etc/pve/nodes/FAILEDHOST/openvz/1xx.conf TO /etc/pve/nodes/LIVEHOST/openvz/1xx.conf and start up the containers AND I manually copy the /private content for that container to the server where I will be restarting it

5. Optionally, if I want to restart all the down containers on the SAME BACKUP HW NODE then I can create a NEW LOCAL DIRECTORY TYPE STORAGE DEVICE---> corresponding to the same ISCSI LUN on a new mountpoint say /data/containers2
and then also edit the *.conf files switching out the storage names.

Is anything missing here? I have tested it and it seems to work....