FastStart doesn't add NC_WORK_SIZE and NC_CACHE_SIZE to eucalyptus.conf

[kamalgill]

Any attempts to launch an instance on a FastStart-installed Eucalyptus cloud returns the following cryptic error in the Eucalyptus Management Console:

Not enough resources (0 in default < 1): vm instances

The /etc/eucalyptus/eucalyptus.conf is:

EUCALYPTUS="/"
LOGLEVEL="INFO"
EUCA_USER="eucalyptus"
CLOUD_OPTS=""
CC_PORT="8774"
SCHEDPOLICY="ROUNDROBIN"
NODES="192.168.1.201"
NC_SERVICE="axis2/services/EucalyptusNC"
NC_PORT="8775"
HYPERVISOR="kvm"
MAX_CORES="0"
INSTANCE_PATH="/var/lib/eucalyptus/instances"
USE_VIRTIO_ROOT="1"
USE_VIRTIO_DISK="1"
USE_VIRTIO_NET="1"
VNET_MODE="EDGE"
VNET_PRIVINTERFACE="br0"
VNET_PUBINTERFACE="br0"
VNET_BRIDGE="br0"
VNET_DHCPDAEMON="/usr/sbin/dhcpd"
METADATA_USE_VM_PRIVATE="N"
DISABLE_TUNNELING="Y"

Note the missing NC_WORK_SIZE setting. Adding NC_WORK_SIZE=20000 allows instances to launch. Not sure if the NC_CACHE_SIZE setting, but NC_WORK_SIZE seems required for a working Eucalyptus cloud.

[lincolnthomas]

The MAX_CORES, NC_WORK_SIZE and NC_CACHE_SIZE settings depend on the user's hardware, so setting a default seems dangerous. But shouldn't we display something very visibly at the end of the FastStart install telling the user to set these values appropriately for their deployment if needed?

The {{MAX_CORES="0"}} basically means no limits. The NC_ settings default to 30% of disk space. So they might not need to be changed at all.

Seems at the very least we should add this to the FastStart doc, @eucalyptus/documentation-team ? Maybe referencing the “Over-subscription” section in the Admin Guide http://docs.hpcloud.com/eucalyptus/4.2.2/#ops-guide/ops_post_install_customizations.html ? (or whatever the link will be for 4.3)

[lincolnthomas]

Clarification: NC_WORK_SIZE defaults to 1/3 of the currently available disk space, and NC_CACHE_SIZE defaults to the other 2/3.