Commit 17a2d789 authored by Leigh B Stoller's avatar Leigh B Stoller
Browse files

Update the installation instructions for the switches. I think I have

everything figured out for setting up the private vlan on both
switches. There is a wire from eth1 on the control node to a port on
the control switch. Another wire from eth2 to a port on the
experimental switch. The complication on the control switch is that we
have no serial console, and so we actually have another wire from eth3
to the control switch, so that we can dhcp and then log in and do a
switcherroo to get the private vlan configured.
parent 0d03e53a
......@@ -70,7 +70,7 @@ Need the following info:
This is about a 1GB so it will take a while.
* Write the image file to the disk using Emulab decompression tool:
* Write the image file to the disk using the Emulab decompression tool:
/usr/bin/imageunzip -o genirack.ndz /dev/cciss/c0d0
......@@ -92,7 +92,7 @@ Need the following info:
have multiple DHCP servers running since that would confuse boss.
Login to the control node as root.
/usr/sbin/dhcpd -4 -cf /etc/dhcp/dhcpd.conf xenbr1 xenbr2
/usr/sbin/dhcpd -4 -cf /etc/dhcp/dhcpd.conf xenbr2 xenbr3
Now power on the experimental switch, and power cycle the control
switch so that it will restart its dhcp cycle. You will lose
......@@ -108,7 +108,29 @@ Need the following info:
to issue the following commands at the prompt.
2610> config
2610(config)> vlan 1 ip address 10.1.1.253/24
2610(config)> vlan 11
2610(vlan-11)> name control-alternate
2610(vlan-11)> untagged 22 XXXX Make sure about port number!
2610(vlan-11)> ip address 10.3.1.253/24
2610(vlan-11)> exit
2610(config)> write memory
Now logout and telnet to 10.3.1.253. This is the alternate link
to the procurve that allows us to configure without a serial link,
and hopefully maintain a connection in case something goes wrong.
More fundimentally, once we remove the ip address from the default
vlan, we would no longer be able to get back in on that IP until
we can recreate it on the private vlan. Basically, we are doing a
switcherroo.
2610> config
2610(config)> no vlan 1 ip address
2610(config)> vlan 10
2610(vlan-10)> name control-hardware
2610(vlan-10)> untagged 21 XXXX Make sure about port number!
2610(vlan-10)> ip address 10.1.1.253/24
2610(vlan-10)> exit
2610(config)> management-vlan 10
2610(config)> ip default-gateway 10.1.1.254
2610(config)> vlan 1 ip igmp
2610(config)> vlan 1 ip igmp querier
......@@ -126,10 +148,25 @@ Need the following info:
The switch will take moment to reset so you might lose your connection
to the control node.
* Now telnet to 10.2.1.253. This is the experimental switch.
* For the experimental switch, we need to do something like above,
which is move the ip address from the default vlan to a private
vlan, but in this case we can do it from the serial console, and
so it is a lot easier. We use minicom:
sudo minicom -D /dev/ttyS0
wait for it to sync up and then you will get the prompt.
5400> config
5400(config)> vlan 1 ip address 10.2.1.253/24
5400(config)> no vlan 1 ip address
5400(config)> vlan 10
5400(vlan-10)> name control-hardware
5400(vlan-10)> untagged 48 XXXX Make sure about port number!
5400(vlan-10)> ip address 10.2.1.253/24
5400(vlan-10)> exit
5400(config)> oobm disable
5400(config)> management-vlan 10
5400(config)> snmp-server listen data
5400(config)> ip default-gateway 10.2.1.254
5400(config)> no web-management
5400(config)> password all (type in same password for manager/operator)
......@@ -158,22 +195,24 @@ Need the following info:
vgcreate xen-vg /dev/sda4
vgchange -a y xen-vg
* Create a small filesystem to hold the boss/ops tarballs. These are
pretty big but will be deleted after we copy the filesystems into
their own lvms.
* Create a filesystem to hold the boss/ops tarballs. These are pretty
big but will be deleted after we copy the filesystems into their own
lvms.
mkdir /scratch
/sbin/lvcreate -n scratch -L 20G xen-vg
/sbin/lvcreate -n scratch -L 50G xen-vg
mke2fs -j /dev/xen-vg/scratch
mount /dev/xen-vg/scratch /scratch
chmod 777 /scratch
* scp boss.tar.gz and ops.tar.gz to to /scratch on the control node,
and then unpack them. There will be two directories, ops and boss.
* Copy boss.tar.gz and ops.tar.gz to /scratch on the control node, and
then unpack them. There will be two directories, ops and boss.
*
* Restore the VMs:
~elabman/clientside/tmcc/linux/xen/restorevm.pl -i boss /scratch/boss
~elabman/clientside/tmcc/linux/xen/restorevm.pl -i ops /scratch/ops
This creates a bunch of LVMs and rewrites the xm.conf in the
boss/ops directories to reflect the new LVM paths, etc.
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment