Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Cool! DRBD over ZVOLs looks like it could be a really sweet, and easier-to-manage HA environment. Thanks for the advise! Miles Simon Ironside wrote: > Hi Miles, > > Yes - I run DRBD inside KVM virtual machines in production just fine. > There were no performance issues or real gotchas that I've encountered. > > I was just careful to not base the VM's storage on DRBD *outside* the > VM in case I accidentally had both VMs running on the same physical > server ;) I was also careful to maintain separate front and back > networks by bridging two separate virtual NICs in the VM to two > separate physical NICs on two separate networks on the physical host > just as I do with DRBD on "real" servers. > > HTH, > Simon. > > On 23/07/14 23:23, Miles Fidelman wrote: >> Hi Folks, >> >> I'm wondering if something is doable - comments please: >> >> I have a very specific situation where I'm trying to maintain a legacy >> high-availability application running in a Xen VM (Linux DomU). >> >> Current stack (per node, in a 2-node HA cluster): >> >> Applications >> VM (Linux DomU) --- auto failover to node 2 ---- >> DRBD ---- replicated to node 2 ---- >> LVM >> md raid >> Xen (Linux Dom0) >> >> >> We're looking at migrating our infrastructure to an OpenSolaris base >> (probably SmartOS), which will end up looking like this: >> >> VM (Linux KVM or Zone) ------------auto failover ot node 2 ----- >> <synchronous replication> ----- to node 2 ---- >> ZFS (exporting a ZVOL) >> KVM (or Zones) >> SmartOS >> >> The kicker, is that there is no available equivalent to DRBD for the >> illumos ecosystem - hence, my thought is to mount the ZVOL block device >> inside the 2 virtual machines, and use DRBD running on Linux inside the >> pair of VMs to achieve storage replication. >> >> Will DRBD run inside a virtual machine? Will it give any kind of >> reasonable performance? Are there any serious gotchas to watch out for? >> >> Three notes re. motivation and thinking: >> - for all our other applications, we're doing application layer >> replication (e.g., database sync) - but for smtp and list processing >> this doesn't seem easily doable (too many queues and messages in flight, >> asynchronously) >> - looking to move from Linux for a number of reasons, but >> particularly ZFS >> - looked at BSD, which has HAST as an equivalent of - virtualization is >> a mixed bag (Xen on NetBSD, but not FreeBSD; KVM sort of, a new >> hypervisor for FreeBSD), and ZFS is only stable on FreeBSD - can do some >> things on FreeBSD, some on NetBSD, but not all on the same one - and >> right now the illumos world seems to be gaining lot of momentum as the >> alternative open-source Unix >> >> Thanks very much, >> >> Miles Fidelman >> >> >> >> >> >> -- In theory, there is no difference between theory and practice. In practice, there is. .... Yogi Berra