Home / user-280505

NStorm's questions

Martin Hope
NStorm
Asked: 2017-09-02 00:24:35 +0800 CST
  • 0

If you are running a 3 or more node cluster with cman/corosync and you have some sort of shared block storage (SAN) connected only to some nodes in the cluster how can you enable CLVM on that storage device?

In my example I'm running a 3 node cluster, there 2 nodes are "main workload" and 1 node are used for backups & archives. Main nodes are connected with FC HBA with multipath to a SAN. Everything works fine, I was able to initialize PV on that device and can see it on both nodes:

  PV                 VG      Fmt  Attr PSize   PFree  
  /dev/mapper/mpathd san-vg1 lvm2 a--  500.00g 500.00g

Now, to create a clustered VG, I've enabled fencing on all 3 nodes in my cluster (with IPMI if that matters).

fence domain
member count  3
victim count  0
victim now    0
master nodeid 1
wait state    none
members       1 2 3

I've also created a clustered VG (vgcreate -c)on that PV and that one works fine too:

  VG      #PV #LV #SN Attr   VSize   VFree  
  san-vg1   1   0   0 wz--nc 500.00g 500.00g

But I cannot create LV on that VG now, because 3rd node in the cluster doesn't have FC HBA, and don't "see" that PV at all:

  Error locking on node main-stor1: Volume group for uuid not found: 3ju5fLfCFyGNivksaq0z7YVGlpWhIX87yK0CoL4e9y86xtwpVDu8PS6WnSdfdlCU
  Failed to activate new LV.

I don't actually require access from that node to VG/LVs stored on that SAN. But I need to have that node synced in the cluster for resource management purposes.

Is there are a way I can tell LVM to ignore/exclude that node from clustered VG?

cluster
Martin Hope
NStorm
Asked: 2015-04-08 21:50:45 +0800 CST
  • 1

I have a node running OpenVZ CT (ProxMox). eth1 on HN are bridged through vmbr1 to veth device on CT101 (eth1 inside container).

(Host Node) eth1 -> vmbr1 <- veth101.1 <- eth1 (CT101)

No VLANs, casual setup. Both bridge and physical interface on HN aren't assigned any IPs. They are just "forwarding" for CT101.

Now I want to forward trunk port with 2 802.1Q VLANs through that interface. Can I just create those VLANs subinterfaces inside CT101? How it works then tags are passed through linux bridge? Are tags are saved or removed?

Or it would be better to detach eth1 from vmbr1 on HN, create VLAN subinteraces and attach them through separate bridges to CT101. So each VLAN goes though dedicated bridges and ends up as an untagged "pseudo-physiacal" ethX device inside container?

linux