Dell PowerEdge 2950 confusion

Dell PowerEdge 2950 confusion
Owner	Space
Status	Infra
Hostname	confusion
Location	Server room
Tool	No
Tool category

Dell PowerEdge 2950 confusion

Confusion.jpg {{{InventoryOwner}}}Property "Tool Owner" (as page type) with input value "{{{InventoryOwner}}}" contains invalid characters or is incomplete and therefore can cause unexpected results during a query or annotation process. Server room

Specs

• 32GB fully buffered DDR2. Max is 32GB or 64GB for gen III
• iDRAC @ [1]
• CPU 2x X5450 Xeon 3GHz (4 cores each)
• HDDs SAS 3x146GB and 3x73GB (264GB logical drive in OS)
• 2 onboard network ports and 2x2 ports on cards.

The owners manual

OS

• Proxmox 4, management on [2]

Network Layout

nurds-sw-04 port 12 set up to accept VLAN tagged packets.
eth2 connected to sw-04 port 23 from upper NIC
eth3 connected to sw-08 port 24 from lower NIC
eth4 connected to sw-08 port 22 from lower NIC
eth5 connected to sw-04 port 12 from upper NIC
bond0 set up but doing nothing on eth[0,1]
bond1 set active-backup on eth[4,2] (cross-card, primary upper)
bond2 set active-backup on eth[3,5] (cross-card, primary lower)
vmbr0 providing management console on VLAN1, bridging bond2 interface. This is for the NFS connections to the storage.
vmbr7 connected to VLAN7, bridging bond1.7 interface (i.e. tagging packets as VLAN7).
vmbr8 connected to VLAN8, bridging bond1.8 interface (i.e. tagging packets as VLAN8).
vmbr7 and vmbr8 available for VM/container binding, with no management interface connection.

Issues

• Wake on LAN on connected NIC does't seem to work. I could not find the appropriate option in the BIOS
• Console redirection in DRAC interface won't work with tried browsers and java

Todo

• DRAC email alerts
• Logins through LDAP

Migrations

As per https://pve.proxmox.com/wiki/Convert_OpenVZ_to_LXC , migration of most of the containers is riduclously easy:

For containers, a script!:

#!/bin/bash
if [ $# -lt 1 ] ; then
    echo "$(basename $0) <NUM to migrate>"
    exit 1
fi
CT=$1
if ssh precious vzlist | grep -F -w -q "  $CT  " ; then
  ssh precious vzctl stop $CT
  ssh precious vzdump $CT -dumpdir /mnt/pve/dumpster-a-vm/dump/
  pct restore $CT /mnt/pve/dumpster-a-vm/dump/vzdump-openvz-${CT}*.tar -storage dumpster-b-vm
  pct set $CT -net0 name=eth0,bridge=vmbr7
  pct start $CT
else
  echo "This container ID doesn't exist on precious"
  exit 1
fi

On a few machines with static IP's, the IP can be entered in the main webconsole.

When sid's been used, proxmox doesn't like it much, so it needs to be switched to 'unmanaged' ostype, e,g,:

pct restore NUM /mnt/pve/dumpster-a-vm/dump/vzdump-openvz-NUM*.tar -storage dumpster-b-vm -ostype unmanaged

For vms:

• on precious:

vzdump NUM -dumpdir /mnt/pve/dumpster-a-vm/dump/

• on confusion:

qmrestore /mnt/pve/dumpster-a-vm/dump/vzdump-qemu-NUM*.vma NUM -storage dumpster-b-vm

DON'T create vms on the same storage block without removing FIRST - it'll duplicate the disks, which is strange when you remove the hosts from the original, and delete the rootfs from under the machine.

Auth

To enable LDAP on this machine, enable it in PAM like on the LDAP page. This does cause auth to request information from one of its own containers, but this is only an issue if the LDAP VM doesn't come up. root is still locally authed.

Then, you MUST restart pvedaemon. This may knock a few containers offline, watch it. Now ldap users exist in realm PAM.

To generate a pool and a group per LDAP user, do this per user:

  pveum useradd "$user"@pam
  pveum groupadd "$user"
  pveum usermod "$user"@pam -append -group "$user"
  pvesh create /pools -poolid "$user"
  pveum aclmod /pool/"$user"/ -group "$user" -role PVEAdmin

VM Issues

Trying to mount nfs inside a proxmox container under LXC won't work, as the default apparmor profile (lxc-container-default-cgns) completely disallows all mount operations. To allow this on specific containers, you need to do this:

• Add mount fstype=nfs to a new apparmor profile (e.g. lxc-default-with-mounting )
• Tell proxmox to use this profile:
  • Edit /etc/pve/lxc/$CTID.conf to say:

lxc.aa_profile = lxc-container-default-with-mounting

• Restart the container.

Trying to use systemd in a container? Then you also need:

fstype=cgroup

In your apparmor profile.

Clustering

This machine is clustered with Precious to allow HA.

VM's

• Graphite.nurdspace.lan
• HackLab
• Revproxy01
• Testwiki