Fedora 18 Single System Install

a13m edited this page Mar 6, 2013 · 5 revisions

Network Configuration

NOTE: This configuration may vary by system, depending on your overall network configuration. If em1 is your only/default interface, it is likely that NetworkManager will drop your connection as soon as ifcfg-em1 changes.

  1. Configure the bridge (so that system mode will work).
cat <<EOF >/etc/sysconfig/network-scripts/ifcfg-br0
DEVICE=br0
ONBOOT=yes
NM_CONTROLLED=no
BOOTPROTO=dhcp
TYPE=Bridge
DELAY=0
PERSISTENT_DHCLIENT=yes
EOF

cat <<EOF >/etc/sysconfig/network-scripts/ifcfg-em1
DEVICE=em1
ONBOOT=yes
NM_CONTROLLED=no
# BOOTPROTO=dhcp
TYPE=Ethernet
BRIDGE=br0
EOF
  1. Fix SELinux context:
chcon -u system_u /etc/sysconfig/network-scripts/ifcfg-*
  1. Enable legacy network setup and remove NetworkManager.
systemctl enable network.service
yum erase NetworkManager NetworkManager-glib
reboot

You could get by without rebooting here, but I find it useful to make sure that the network comes up properly after a reboot before proceeding with the rest of the configuration. If you do not want to reboot, you need to do this before removing NetworkManager:

systemctl stop NetworkManager; systemctl start network

Kernel Parameter Configuration

  1. These kernel parameters are required for postgresql to function properly. This was addressed in one way in EUCA-2034, but that solution is not acceptable in Fedora, so it must be done manually.
cat <<EOF >/etc/sysctl.d/eucalyptus-cloud
kernel.shmmax = 536870912
kernel.sem = 250 32000 32 32768
EOF
sysctl -p /etc/sysctl.d/eucalyptus-cloud

Eucalyptus Package Installation

yum install eucalyptus-cloud eucalyptus-sc eucalyptus-walrus eucalyptus-cc eucalyptus-nc

Cloud Initialization

su eucalyptus -c "euca_conf --initialize"
systemctl start eucalyptus-cloud

The first startup can take a few minutes while the databases are loaded. You can tail -f /var/log/eucalyptus/cloud-exhaust.log and watch for the "Mule ESB and Integration Platform" box to appear, which is a good indication that the web services stack is fully loaded.

Component Registration

  • Get credentials - Note that if you don't wait long enough, you may get an error or a zero-length file here. If so, remove the empty file and try again.
euca_conf --get-credentials admin.zip
unzip admin.zip
source ./eucarc
  • Verify that the java-based services are all in ENABLED state (except for the Dns service, which will likely be "broken" due to conflict with dnsmasq):
euca-describe-services
  • Register and start the remaining components
euca_conf --register-walrus -C walrus -P walrus -H <host-ip-address>
euca_conf --register-sc -C SC00 -P PARTI00 -H <host-ip-address>
euca-modify-property -p PARTI00.storage.blockstoragemanager=overlay
euca_conf --register-cluster -C CC00 -P PARTI00 -H <host-ip-address>
systemctl start eucalyptus-cc
euca_conf --register-nodes <host-ip-address>
systemctl start eucalyptus-nc
systemctl start eucalyptus-console

Verify compute resources

At this point, if all services are working properly, you should be able to see available resources:

euca-describe-availability-zones verbose

If you do not, it is likely that either the CC or NC has not started properly. Here are a couple of known issues:

  • If you do not reboot the system or reload the kvm modules after installing the qemu packages, then /dev/kvm may have the wrong permissions. You can fix this by reloading the modules:
rmmod kvm_intel
rmmod kvm
modprobe kvm_intel

Image Registration

wget -nd http://s3.amazonaws.com/fedora17-201209050306/initramfs-3.5.3-1.fc17.x86_64.img
wget -nd http://s3.amazonaws.com/fedora17-201209050306/ks-fedora17-201209050306.img.gz
wget -nd http://s3.amazonaws.com/fedora17-201209050306/vmlinuz-3.5.3-1.fc17.x86_64
euca-bundle-image  --kernel true -i vmlinuz-3.5.3-1.fc17.x86_64
euca-upload-bundle -m /tmp/vmlinuz-3.5.3-1.fc17.x86_64.manifest.xml -b fedora17
EKI_ID=$( euca-register fedora17/vmlinuz-3.5.3-1.fc17.x86_64.manifest.xml )
euca-bundle-image --ramdisk true -i initramfs-3.5.3-1.fc17.x86_64.img 
euca-upload-bundle -m /tmp/initramfs-3.5.3-1.fc17.x86_64.img.manifest.xml -b fedora17
ERI_ID=$( euca-register fedora17/initramfs-3.5.3-1.fc17.x86_64.img.manifest.xml )
gunzip ks-fedora17-201209050306.img.gz; 
euca-bundle-image --kernel $EKI_ID --ramdisk $ERI_ID -i ks-fedora17-201209050306.img
euca-upload-bundle -m /tmp/ks-fedora17-201209050306.img.manifest.xml -b fedora17
EMI_ID=$( euca-register fedora17/ks-fedora17-201209050306.img.manifest.xml )

Test Instance Launch

euca-add-keypair foobar > foobar.priv
chmod 600 foobar.priv
euca-run-instances -k foobar -t m1.large $EMI_ID

Note: The instance boot time is really slow, partly due to SELinux file relabelling. Also, since I'm testing system mode here, cloud-init fails, so you have to log in as root rather than ec2-user. It is possible to set up metadata service access for system mode, but I haven't done the iptables rules for that yet.


category.install

Clone this wiki locally
You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.
Press h to open a hovercard with more details.