request to Proxmox team: Best support of last version of DRBD

[cesarpk]

Congratulations for Proxmox team, proxmox is a product wonderful !!!
champagne, champagne !!!

Now I would like to make a request to the Proxmox team:

Considering that "Proxmox" currently support High Availability,
Is it possible that the repository of "Proxmox" has the latest version of "DRDB" and kernel supported?
(actually DRBD version is 8.4.1 and not 8.3.10, and many bugs have been fixed)

- I think it will be a very important advance, because a vast majority of users will not need (or can not buy) double external storage for HA in storage.
- Also for the case of storage, stability plays a major factor in information security.

Best Regards and congratulations again
Cesar

 

[e100]

Proxmox uses the kernel from the openvz project which is why I filed a bug with openvz requesting a newer version:
http://forum.proxmox.com/threads/93...lready-patched-by-DRBD-devs?p=53118#post53118

http://bugzilla.openvz.org/show_bug.cgi?id=2254

I have 14 Proxmox servers running DRBD some running over two years now, I only have had one bug bother me (causes no data loss nor downtime) and it was just fixed last month by DRBD folks.
Hopefully we will hear from the openvz devs soon.

 

[cesarpk]

Thank you e100, I feel more relaxed knowing that no big problem, I hope that openvz project soon fix the bugs.

But, do you have Proxmox his own kernel for support openvz and kvm based on debian?

And in this link: http://martinjohnstone.com/2012/03/20/proxmox-2-and-drbd-8-4/ , i see it is possible to prepare "Proxmox" for use with the latest version of "drbd", including the kernel.

What do you think about that?

Best regards
Cesar

 

[e100]

Proxmox uses the openvz kernel that supports kvm and openvz.

Yes, it is simple to compile the DRBD userland and kernel module yourself. I prefer not to use that method because it would be more work when updates are installed.

If it takes too long to get the DRBD module updated I might decide to compile it myself.

The only bug that I have seen has been around for years and only recently identified and fixed. Waiting another month or so to see it fixed in Proxmox is not going to bother me too much.

I have been running 2.0 since it was first released in beta. Two servers have been running production VMs for about 6 weeks now. We have seen the DRBD bug once.

It causes a Protocol error resulting in a split brain so replication gets interrupted.
Our monitoring system alerts us, we log in run a few commands and DRBD is repaired.

Yes it needs fixed, but it is not a critical issue that needs fixed imediatly.

 

[cesarpk]

Thank you very much e100 for your prompt response and time for explain me

I was thinking to use proxmox only on two nodes + "HA" + KVM (only) + qcow2 + drbd + gfs2 + differents VMs on both nodes, this idea because it makes better use of processor resources. Now I have doubts and believe that better is only run several VMs only on one node and the second proxmox node will be passive node. What is your opinion?

Best regards
Cesar

 

[udo]

Thank you very much e100 for your prompt response and time for explain me

I was thinking to use proxmox only on two nodes + "HA" + KVM (only) + qcow2 + drbd + gfs2 + differents VMs on both nodes, this idea because it makes better use of processor resources. Now I have doubts and believe that better is only run several VMs only on one node and the second proxmox node will be passive node. What is your opinion?

Best regards
Cesar

Hi,
for HA you should think about a third node... with two nodes it's not easy to get an solid working solution. gfs2 (and qcow2 on this) is for me not easy enough - esp. in this small setup. DRBD and lvm ontop work for me very well (at this time on pve1.9 - and on one 3node testsystem with pve2). Soon i will build an 3node pve2-cluster with DRBD on two nodes (but without HA in the beginning).

As you only want to use kvm, you don't need an filesystem (only for backup and isos) - so lvm-storage seems to be the right thing for you.

Udo

 

[cesarpk]

Thank you udo for your response

Let me ask another question:

I need configure proxmox VE 2.x with:
- Only two nodes (I know configure proxmox in cluster (not to be confused with Cluster HA))
- DRBD ( I know DRBD very well for configure primary/primary and primary/secundary)
- Extra disk for DRBD in each node
- KVM (not openvz)
- Without fence devices
- Without storage external (NFS, iSCSI, etc.)

The idea is that if i have 2 nodes with VMs are powered on, and in case of breakdown of a node, manually start the VMs on the other node

And I need to know:
- How configure Proxmox for do it?

References:
- Because there is no tutorials for this case
- And neither for use with only 2 nodes:
On the website:
http://pve.proxmox.com/wiki/Two-Node...bility_Cluster
literally says: "For this testing configuration, two DRBD resources were created, one for VM images an another one for VMs users data", but does not say how to do it
And on the website:
http://pve.proxmox.com/wiki/DRBD
no says how configure VMs users data
Then, there is no tutorials for 2 nodes

- For my test, when I turn off a Proxmox Cluster node, the other Proxmox Cluster node is automatically disabled, and can not be used it the second node until it is restarted, so it is impossible to test for faults.

can you help me?

Best regards
Cesar

 

[udo]

Thank you udo for your response

Let me ask another question:

I need configure proxmox VE 2.x with:
- Only two nodes (I know configure proxmox in cluster (not to be confused with Cluster HA))
- DRBD ( I know DRBD very well for configure primary/primary and primary/secundary)
- Extra disk for DRBD in each node
- KVM (not openvz)
- Without fence devices
- Without storage external (NFS, iSCSI, etc.)

The idea is that if i have 2 nodes with VMs are powered on, and in case of breakdown of a node, manually start the VMs on the other node

And I need to know:
- How configure Proxmox for do it?

References:
- Because there is no tutorials for this case
- And neither for use with only 2 nodes:
On the website:
http://pve.proxmox.com/wiki/Two-Node...bility_Cluster
literally says: "For this testing configuration, two DRBD resources were created, one for VM images an another one for VMs users data", but does not say how to do it
And on the website:
http://pve.proxmox.com/wiki/DRBD
no says how configure VMs users data
Then, there is no tutorials for 2 nodes

- For my test, when I turn off a Proxmox Cluster node, the other Proxmox Cluster node is automatically disabled, and can not be used it the second node until it is restarted, so it is impossible to test for faults.

can you help me?

Best regards
Cesar

Only for references: this is answered here http://forum.proxmox.com/threads/94...entation-on-Proxmox-website?p=53249#post53249

 

[rpotter28]

This thread demonstrates the power of this forum, Open Source and Proxmox with the continuous contributions of Proxmox developers/staff and the 2 users in this thread being udo and e100. I am a seasoned 20 year Linux guy, a Red Hat Certified Engineer and I learn things here. This is a very good project and deserves to be supported. i will be doing so. udo and e100 deserve recognition for their efforts here, and I am doing so. Thanks!

 

[cesarpk]

This thread demonstrates the power of this forum, Open Source and Proxmox with the continuous contributions of Proxmox developers/staff and the 2 users in this thread being udo and e100. I am a seasoned 20 year Linux guy, a Red Hat Certified Engineer and I learn things here. This is a very good project and deserves to be supported. i will be doing so. udo and e100 deserve recognition for their efforts here, and I am doing so. Thanks!

I totally agree rpotter28, udo and e100 deserve recognition for their efforts here, I acknowledge the effort of this team that is wonderful !!!.

Thanks Proxmox team !!!!

 

[hrodriguez231]

I have been running 2.0 since it was first released in beta. Two servers have been running production VMs for about 6 weeks now. We have seen the DRBD bug once.

It causes a Protocol error resulting in a split brain so replication gets interrupted.
Our monitoring system alerts us, we log in run a few commands and DRBD is repaired.

Hi e100 and Udo, I have read some other posts of you two telling that you have configured servers using DRBD without problems for many months, could you give us step by step how to configure 2 servers with DRBD, and possible include in the future 2 or 3 more servers in the clusters, but also having DRBD.
what are your suggestions in this regard?

Could you tell exactly the configuration you have done? In my case I have 1 HP Proliant server, and 3 workstation with Proxmox installed, I would like to know whether to configure DRBD only 2 servers at the same time, and take another 2 and do the same.
I have no SAN or NAS, each workstation has 500GB hdd, 4GB RAM DDR3, Core i3.


I will use KVM and CT, but mostly KVM.
Thanks in advance,
Hector.

 

[udo]

Hi e100 and Udo, I have read some other posts of you two telling that you have configured servers using DRBD without problems for many months, could you give us step by step how to configure 2 servers with DRBD, and possible include in the future 2 or 3 more servers in the clusters, but also having DRBD.
what are your suggestions in this regard?

Could you tell exactly the configuration you have done? In my case I have 1 HP Proliant server, and 3 workstation with Proxmox installed, I would like to know whether to configure DRBD only 2 servers at the same time, and take another 2 and do the same.
I have no SAN or NAS, each workstation has 500GB hdd, 4GB RAM DDR3, Core i3.


I will use KVM and CT, but mostly KVM.
Thanks in advance,
Hector.

Hi Hector,
I use drbd only for kvm - for openvz you need an cluster-filesystem (or an nfs-server which run as ha with an extra drbd-setup).

But I don't know if you will be happy with your setup - you need io-power, and with only one sata-drive you don't get good io-values.

Typical setup for me: raid-10 (sas or sata), with fast raid-controller (e.g. areca 1222), two raid-volumes (or one volume with two partitions. one for each server).
On the raid-disk(s)/partition an drbd-resource and on the resource an volumegroup. This volume-group must be defined in the pve-gui.
Other things like in the drbd-wiki-entry.
The connection between the both nodes with 10GB-Ethernet, dolphin-adapter or infiniband.

Udo

 

[e100]

I pretty much followed the wiki when setting up DRBD except I create two DRBD volumes(which it suggests just is not included in the main directions)

Udo mentioned using a RAID card, I agree, it makes a huge difference in performance with DRBD.
SSD for DRBD disks would also work well too I suspect.

If your IO needs are modest a single 1G Ethernet port will limit you to about 100MB/sec write speed.
If you need something faster used 10G Infiniband gear is very cheap, that is what I am using.

If you want to use DRBD I suggest building them two at a time.
You can have multiple two node DRBD setups in one Proxmox cluster.