Difference between revisions of "DRBD9"
|Line 246:||Line 246:|
== Recovery from communication failure ==
== Recovery from communication failure ==
== Integrity checking ==
== Integrity checking ==
Revision as of 13:40, 17 January 2016
DRBD® refers to block devices designed as a building block to form high availability (HA) clusters. This is done by mirroring a whole block device via an assigned network. DRBD can be understood as network based raid-1. For detailed information please visit Linbit.
Main features of the integration in Proxmox VE:
- drbd9/drbdmanage; drbd devices on top of LVM
- All VM disks (LVM volumes on the DRBD device) can be replicated in real time on several Proxmox VE nodes via the network.
- Ability to live migrate running machines without downtime in a few seconds WITHOUT the need of SAN (iSCSI, FC, NFS) as the data is already on both nodes.
- LXC containers can use DRBD9 storage
DRBD9 integration is introduced in Proxmox VE 4.0 as technology preview.
You need 3 identical Proxmox VE servers (V4.0 or higher) with the following extra hardware:
- Extra NIC (dedicated for DRBD traffic)
- Second disk, SSD, Flash card or raid volume (e.g. /dev/sdb) for DRBD
- Use a hardware raid controller with BBU to eliminate performance issues concerning internal metadata (see Florian´s blog).
- A functional Proxmox VE Cluster (V4.0 or higher)
- At least 2GB RAM in each node
VM settings when running on top of DRBD
- DRBD supports only the raw disk format at the moment.
- You need to change the VM disk cache mode from the default 'none' to 'writethrough' instead of the default 'none'. Do not use write cache for any virtual drives on top of DRBD as this can cause out of sync blocks. Follow the link for more information: http://forum.proxmox.com/threads/18259-KVM-on-top-of-DRBD-and-out-of-sync-long-term-investigation-results?p=93126
- Consider doing integrity checking periodically to be sure DRBD is consistent
Configure the NIC dedicated for DRBD traffic (eth1 in the current example) on all nodes with a fixed private IP address via the web interface and reboot each server.
For better understanding, here is an /etc/network/interfaces example from the first node called pve1, after the reboot:
cat /etc/network/interfaces # network interface settings auto lo iface lo inet loopback iface eth0 inet manual auto eth1 iface eth1 inet static address 10.0.15.81 netmask 255.255.255.0 auto vmbr0 iface vmbr0 inet static address 192.168.15.81 netmask 255.255.255.0 gateway 192.168.15.1 bridge_ports eth0 bridge_stp off bridge_fd 0
And from the second node, called pve2:
# network interface settings auto lo iface lo inet loopback iface eth0 inet manual auto eth1 iface eth1 inet static address 10.0.15.82 netmask 255.255.240.0 auto vmbr0 iface vmbr0 inet static address 192.168.15.82 netmask 255.255.255.0 gateway 192.168.15.1 bridge_ports eth0 bridge_stp off bridge_fd 0
And finally from the third node pve3:
# network interface settings auto lo iface lo inet loopback iface eth0 inet manual auto eth1 iface eth1 inet static address 10.0.0.83 netmask 255.255.255.0 auto vmbr0 iface vmbr0 inet static address 192.168.15.83 netmask 255.255.255.0 gateway 192.168.15.1 bridge_ports eth0 bridge_stp off bridge_fd 0
Disk for DRBD
DRBD will search for the LVM Volume Group drbdpool. So you have to create them on all nodes.
I will use /dev/sdb1 for DRBD. Therefore I need to create this single big partition on /dev/sdb - make sure they exist on all nodes.
To prepare the disk for DRBD just run
parted /dev/sdb mktable gpt parted /dev/sdb mkpart drbd 1 100% parted /dev/sdb p Model: ATA Samsung SSD 850 (scsi) Disk /dev/sdb: 512GB Sector size (logical/physical): 512B/512B Partition Table: gpt Disk Flags: Number Start End Size File system Name Flags 1 1049kB 512GB 512GB drbd
And create then logical volume dedicated for drbd
NOTE: The logical volumes must have all the same size on each node!
root@proxmox:~# vgcreate drbdpool /dev/sdb1 Physical volume "/dev/sdb1" successfully created Volume group "drbdpool" successfully created root@proxmox:~# lvcreate -L 511G -n drbdthinpool -T drbdpool Logical volume "drbdthinpool" created.
Remember to leave at least 1 extent available to the volume group, if you don't, drdbmanage will fail with the following error : "Volume group "drbdpool" has insufficient free space (0 extents): 1 required."
As LVM thin provisionning is used at the logical volume level, install the following package :
apt-get install thin-provisioning-tools
Install the DRBD user tools on all nodes :
apt-get install drbdmanage -y
First make sure that the ssh-keys of each node are in "known_hosts" list from all the other. This can be easily ensured by
root@pve1:~# ssh 10.0.15.82 root@pve1:~# ssh 10.0.15.83
root@pve2:~# ssh 10.0.15.81 root@pve2:~# ssh 10.0.15.83
root@pve3:~# ssh 10.0.15.81 root@pve3:~# ssh 10.0.15.82
To configure DRBD9 it is only necessary to run the following command on node pve1:
drbdmanage init -q 10.0.15.81 Failed to find logical volume "drbdpool/.drbdctrl_0" Failed to find logical volume "drbdpool/.drbdctrl_1" Logical volume ".drbdctrl_0" created. Logical volume ".drbdctrl_1" created. initializing activity log NOT initializing bitmap Writing meta data... New drbd meta data block successfully created. initializing activity log NOT initializing bitmap Writing meta data... New drbd meta data block successfully created. empty drbdmanage control volume initialized. empty drbdmanage control volume initialized. Operation completed successfully
Now add all nodes of the cluster to DRBD, with the following commands, still on node pve1 (you should check ssh login as root to these nodes work)
root@pve1:~# drbdmanage add-node -q pve2 10.0.15.82 Operation completed successfully Operation completed successfully Executing join command using ssh. IMPORTANT: The output you see comes from pve2 IMPORTANT: Your input is executed on pve2 Failed to find logical volume "drbdpool/.drbdctrl_0" Failed to find logical volume "drbdpool/.drbdctrl_1" Logical volume ".drbdctrl_0" created. Logical volume ".drbdctrl_1" created. NOT initializing bitmap initializing activity log Writing meta data... New drbd meta data block successfully created. NOT initializing bitmap initializing activity log Writing meta data... New drbd meta data block successfully created. Operation completed successfully
and then finally
root@pve1:~# drbdmanage add-node -q pve3 10.0.15.83 [...]
then add a DRBD entry to /etc/pve/storage.cfg like this:
NOTE1: Redundancy <Number> - this number can not be higher than the maximum number of your actual total nodes.
NOTE2: If the file does not exist, try adding some storage in web GUI like a "local directory" one, and pve will create the file for you
NOTE3: Each storage entry in that file must be followed by exactly one empty line
drbd: drbd1 content images,rootdir redundancy 3
The node configuration can be verified by
Create the first VM on DRBD for testing and live migration
On the GUI you can see the DRBD storage and you can use it for as virtual disk storage.
NOTE: DRBD supports only raw disk format at the moment.
Try to live migrate the VM - as all data is available on both nodes it will take just a few seconds. The overall process might take a bit longer if the VM is under load and if there is a lot of RAM involved. But in any case, the downtime is minimal and you will see no interruption at all.
DRBD can be configured in many different ways and there is a lot of space for optimizations and performance tuning. If you run DRBD in a production environment we highly recommend the DRBD commercial support from the DRBD developers. The company behind DRBD is Linbit.
Recovery from communication failure
If the communication between storage is interrupted but the nic is still up, when node reconnects they will sync again.
If the nic instead goes down (i.e. cable unplugged), then when the nic and communication is up again the nodes remain isolated. For example if you have prox01 with resource A as primary and prox02, when you reconnect you will see (# drbdsetup status) on prox01 "prox02 connection:Connecting" and in prox02 "prox01 connection:StandAlone" and it's disks flagged as "outdated". When in "StandAlone" state no sync it automatically performed. To force the reconnection you have to issue on node prox02 the following command:
root@prox02:~# drbdadm adjust all
The same applies if you create a resource on a node when it's disconnected from the others.
- You can enable "data-integrity-alg" for testing purposes and test at least for a week before production use. Don't use in production as this can cause split brain in dual-primary configuration and also it decreases performance.
- It is good idea to run "drbdadm verify" once a week (or at least once a month) when servers under low load.
# /etc/cron.d/drbdadm-verify-weekly # This will have cron invoked a drbd resources verification every Monday at 42 minutes past midnight 42 0 * * 1 root /sbin/drbdadm verify all
- Check man drbd.conf, section "NOTES ON DATA INTEGRITY" for more information.
Now you have a fully redundant storage for your VM´s without using expensive SAN equipment, configured in about 10 to 30 minutes - starting from bare-metal.
- Three servers for a redundant SAN
- Three servers for redundant virtualization hosts
Alternative Two storage node setup
You can also setup a cluster with 2 powerful servers that has storage, and a third node with a lightweight PC (i.e. Mitac Pluto 220) just for quorum. The difference are (if 'pve3', IP 192.168.15.83, is the node without DRBD storage):
- In node pve3 you don't need to configure eth1 (was 10.0.0.83) for DRBD storage communication
- In node pve3 you don't have to configure storage nor DRBD
- From pve1 you don't have to # drbdmanage add-node -q pve3 10.0.15.83
- In storage.cfg you will have "redundancy 2" instead of 3 and you should add a line like this to limit the DRBD storage visibility: nodes pve1,pve2
The only drawback is that node 3 is listed when you choose target for migration but is not a good choice!