All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Proxmox OpenVZ container question
I have used OVH's dedicated servers in the past. Recently I ordered an online.net dedi, and installed Proxmox 2.3 on it, upgraded to Proxmox 3.
I ordered an IPv4 failover for using an openvz container. I added the IPv4 to the server via online.net's panel, created a new container in Proxmox added the IP in "routed mode", and started the container. Now in OVH, that was all that was needed to connect to the container. However I noticed that the new IP was not pingable.
I then added the following to /etc/network/interfaces:
#Failover configuration auto dummy0 iface dummy0 inet static address 6.6.6.6 netmask 255.255.255.255
where 6.6.6.6 is my failover IP.
The IP became pingable, however, instead of connecting to the container, the failover IP connects to the main server. How does one configure it to get the failover IP to connect to the container?
Comments
Try seeing if it works in bridged mode
Apparently the default configuration is bridged mode on online.net.
You tried putting the container in bridged mode?
I made a mess of /etc/network/interfaces, and couldnt connect to the node. Reinstalling now..
Ok, the node was already in bridge mode when installed. How should I be configuring it for routed (or bridge) mode?
@joelgm - I think that needs to be a bridge not dummy0 in the config, as dummy0 would bridge the main IP/interface
Failover configuration
auto vmbr2
iface vmbr2 inet static
address 6.6.6.6
netmask 255.255.255.255
bridge_ports eth0:0 (or whatever interface you assigned the IP to on the server)
bridge_stp off
bridge_fd 0
gateway ?.?.?.?
then in proxmox select vmbr2 instead of vmbr0 for the bridge to eth0 of the container (bridge mode).
Please note: I am not an expert
Under "just trying to help":
I normally do not need the extra server config and can just add the IP to the container (routed mode) , without adding the IP to the server at all.
If you want to mess with dummys without killing the main IP,
Add an extra dummy
echo "options dummy numdummies=2" > /etc/modprobe.d/local.conf (restart)
Then after restart check that the extra dummy appeared
ls /proc/sys/net/ipv4/conf/
You should now have
all dummy0 eth0 venet0 default dummy1 lo vmbr0
if the dummys do not show, don't try and bridge to it.
Edit: changed bridged to routed in 2nd line
Have you removed the IP binding from the node, and add it only to the container?
Don't make any configuration in the networking, just try it first. I'm trying in the Online.net in the past, and it can work.
Er.. How do I do that?
I tried doing no network configuration at all, and creating a container by simply assigning IP (routed mode) as I've done in the past for OVH servers. However it was not connecting to the container at all.
I set it up as you kindly advised. I edited the code you provided to include my failover IP, rebooted, created a container in bridge mode and chose vmbr2
My network config is: http://www.pastebin.ca/2468202
I then created a container in bridge mode, and assigned vmbr0. Now it simply connects to the node, when I try to ssh into the failover ip.
Edit: It stopped working again.
How about this?
Do you use the IPv4 configuration same like you get from Online.net for the netmask, network, broadcast and gateway?
Try to use only that line, do not add another IP in the interfaces, but use the failover IP at the container only.
When you restarted/reboot, and connecting to the failover IP, it should connected to the container.
Previously, I only have one IPv4 and connecting to the Online.net container via port routing. And using the above interface setup for my dedi from FitVPS
@joelgm - I am confused by your setting up two bridged containers and want to make sure I understand what you are trying to do.
1 - You have a service on port XX on the node at the primary IP. The failover IP is set up on one container that runs a duplicate service at port XX. You stop the service on the primary, the service then failsover to the container and the service continues uninterrupted. (one primary IP at the node, one failover IP at the container and you are good to go)
2 - You are trying to test failover between two containers on the same node and is the reason you made two containers above. You have a service on port XX on the container at the primary IP. The failover IP is set up on the secondary container that runs a duplicate service at port XX. You stop the service on the primary container, the service then failsover to the secondary container and the service continues uninterrupted. (You need a third IP for the node not related to failover between the containers)
3 - something else
@FrankZ What I intended to do is set up a master server or node for me to administer, and seperate openvz containers which may be better termed clients, which run independent vps boxes. In short, I need a node and some vps es. All of them would be active at the same time. I shouldn't have to stop a service to run one on a container. Something akin to what solus master and slaves do?
@joelgm - I was good up until "Something akin to what solus master and slaves do?".
Which I take to mean a proxmox HA cluster.
If you are trying to get a couple of openVZ containers "clients" that you can control and access running on a single proxmox node .
Using the config above with vmbr0 and vmbr2 what happens when you only start one container/instance on vmbr2 and no containers using vmbr0?
You should be able to ssh to the node using the vmbr0 IP and ssh to the container using the vmbr2 IP.
Does it happen?
EDIT: Structure
Lets do this by PM to stop bumping the thread
Just an update to note that FrankZ solved the issue. My failover IPs were in a different subnet and required an additional setting in /etc/vz/vz.conf. Thanks a lot, FrankZ!
I wish joelm would have posted what the solution was since I am struggling for 2 days with exactly the same issue. On OVH machines, you install Proxmox template and then simply enter the failover IP in the new openVZ container as routed and you are done.
Online.net phone help aint much help, they dont speak english so only words he said is ticket ticket.
Ok I created a ticket and the result was a link to a FRENCH explanaition website and it was intent for normal Linux distro's and not Proxmox VE 3.
If somebody of online.net is reading this also, please add a Proxmox template which uses routed and not bridged.
Anyway, I have tried everything and I am lost, the node was on vmbr0 and I changed that to eth0 and rebooted. The node is indeed back online, but container retains unreachable.
I also noticed that online.net says I should use submask 255.255.255.255 but Proxmox gives an error on that.
Any help would be very usefull here.
Greetings,
Rev
@Revdutchie:
Did you try subnet 255.255.255.248?
@Revdutchie, I've detailed the modifications needed, on my blog - http://blog.droidzone.in/2013/10/20/setting-up-a-dedicated-server-on-online-net/
joelgm,
So you are saying that if I do a fresh Proxmox install the only thing I need to do is to change
1
NEIGHBOUR_DEVS=detect
to:
?
1
NEIGHBOUR_DEVS=all
Yup
Reinstalling at the moment, keeping my fingers crossed
joelgm you are a LEGEND (for me) but still think you should have posted the solution here when you found it.
But that brings me to Online support. If it is this simple then
A) Why is the phone support that bad, and also ticket support. They should know this.
As I mentioned above, why dont they have 2 templates. One bridged, mostly used for XEN or KVM solutions. And one template routed for openVZ.
Anyway, I am happy afetr 2 days of struggling I got it working.
Cheers,
Rev
@Revdutchie - This is only required if your IPs are on different subnets.
For others future reference the command to run is:
sed -i 's/NEIGHBOUR_DEVS=detect/NEIGHBOUR_DEVS=all/g' /etc/vz/vz.conf
@FrankZ
Nice work Zappa.. thank god I was too cheap to buy a failover IP from online.net or I would probably be scratching my head too. lol
Anyone that has more knowledge then i, is an expert..
Maybe a dumb question, but could the topic be changed with something like online.net servers in the name so our future googlers can find it? online.net is a great provider name but a pain when you gonna search google with online in the search string :-)
If you search for
"online.net" dedicated proxmox
, my blog post gets the second hit. To rename the title on this thread, you'd need to contact a mod. I don't have permission to do it myself.