gc3-uzh-ch / openstack-tools

openstack-tools

This is a collection of tools useful to manage/inspect the status of an OpenStack installation.

Please note that these tools have been tested only on Ubuntu 12.04, OpenStack Folsom!

create_image.sh

This tool is used to create and possibly upload to a glance instance a virtual cloud-ready virtual disk. The tool allow you to create virtual machines for one of the following distributions:

• CentOS 5, 6
• Ubuntu Precise Pangolin 12.04 LTS
• Ubuntu Quantal Quetzal 12.10
• Ubuntu Raring Ringtail 13.04
• Ubuntu Trusty Tahr Ringtail 14.04 LTS
• Debian Wheezy 7

In order to do this, it uses virt-install command and a kickstart or a preseed, depending if you are installing a CentOS or a Debian/Ubuntu system. Predefined kickstart and preseeds are stored in the etc/ directory, but you can customize them or supply your own file.

The following packages are required:

• virtinst (virt-install command)
• libguestfs-tools (virt-sysprep and virt-sparsify commands)
• python-glanceclient (glance command, required only to upload the image to OpenStack)

Customization performed on the created images consist of:

• install cloud-init and cloud-initramfs-growroot
• create a gc3-user user (instead of the distro default) and configure cloud-init accordingly
• ensure the cloud-init datasource include EC2
• enable serial console (useful to show console logs from OpenStack)
• ensure the system has ssh host keys
• configure via dhcp any network interface (not just the first one)

Tools present in the bin directory

openstack-accounting-usage

A script to read out the OpenStack mysql database and create a report about walltime and volume usage on a per user base.

To use this script you should create a mysql user that has read-only rights.

openstack-check-quotas

Check that nova and cinder quotas in OpenStack reflect the actual usage of the cloud. This is done by checking the quota information stored on the database against the list of running VMs.

An option --update-usages allows you to also fix the quota tables in the DB by updating the information on the actual usage, based on the real actual usage.

If option -k is given, then user and tenant IDs are translated into their names using Keystone. This also requires that you pass --os-username, --os-password, --os-tenant-name and --os-auth-url options, or you set the corresponding environment variable before calling this command.

The script will look for configuration files /etc/nova/nova.conf and /etc/cinder/cinder.conf, in order to get the sql_connection strings and access the nova and cinder databases. You can specify different path for the configuration files using --nova-config-file and --cinder-config-file, or you can specify a SQLAlchemy-like connection string using --nova-sql-string and --cinder-sql-string.

openstack-check-spurious-vms

This tool will check if there are Virtual Machines running on some compute node which are not related to any OpenStack instance.

If -k option is given, then these spurious machines are terminated (by issuing virsh destroy on the compute node)

In order to make this command work you need to:

• supply a valid /etc/nova/nova.conf. This tool uses the same APIs used by the nova-manage tools. As such, you will need to provide a proper /etc/nova/nova.conf in order to be able to access the database.

• ensure that your user can ssh without password on all the compute nodes.

• ensure that virsh is installed on the compute nodes. This also mean that the command will only work if you have virsh installed on the compute node and if libvirt.LibvirtDriver is used as compute_driver.

openstack-create-delete-user

This command line tool automates the creation and deletion of an user in Keystone. It automatically:

• creates/delete the user
• associate the user to the requested role (by default: Member)
• create the tenant if it doesn't exist (and --create-project option is given)
• set rules for the default security group in case a new tenant is created
• remembers you to add the user to the cloud mailing list (very GC3-specific!)

openstack-free-usage

This tool is used to check how many virtual machines your OpenStack can still run for each flavor. Also, it allow you to simulate the existence of a flavor defined using command line options.

By default, a list of flavor is printed to standard output, and two numbers for each flavor next to it is displayed: the number of virtual machines that can be started using that flavor considering the current usage, and the maximum number of instances with that flavor that your cloud can run.

Option -f allows you to specify a list of flavors. If this option is given, only these flavors will be considered and printed

Please note that this tool also checks the extra specs and confront them with the aggregates. This behavior reflects your installation only if your nova.conf defines the AggregateInstanceExtraSpecsFilter filter:

scheduler_default_filters=AggregateInstanceExtraSpecsFilter,...

Option --simulate`` instead will allow you to define a new flavor (using options -n, -c, -m, -r, -eand--extra-specs` to define the name, the number of cpus, the amount of ram, root disk, ephemeral disk and extra specifications) and to check also against this fake flavor. This is very useful to check how many virtual machines your cloud will be able to run with a specific flavor you want to create before actually creating it.

By adding multiple -v options you will get a more verbose output:

-v also print the number of compute nodes, total number of CPUs, CPUs currently in use, total ram, total ram in use, total disk, total disk in use.

-vv also print information about each flavor (vcus, ram, disk etc)

-vvv also print one line for each compute node, with the number of free/total cpus, free/total ram, free/total disk

This tool uses the same APIs used by the nova-manage tools. As such, you will need to provide a proper /etc/nova/nova.conf in order to be able to access the database.

openstack-get-vm-from-ip

This tool access the database and reports all the virtual machines that ever used a specific IP address.

Optionally you can specify a start and end date, to select only the instances that were running within that time range.

If you provide valid OpenStack credentials (either with --os-* options or setting the corresponding OS_* environment variables) and supply the -k option, project and user IDs are converted into project and user names.

Please note that this command only work with fixed IPs or automatically assigned floating IPs, it does not work with "proper" floating IPs, because the required information are not stored in the database.

openstack-instances-by-users

This tool will print all the instances currently running grouped by user. For each instance a few information are printed (name, nr. of cpus, flavor, hostname of the compute node, when has been created, status and the project name). It will also print how many VMs and vCPUs an user is currently using.

You can show information about only a few users with the -u option, followed by the list of users you are interested to.

You can sort the output based on user name, number of cpus used or number of vms currently running (default: sort by name)

This tool uses the same APIs used by the nova-manage tools. As such, you will need to provide a proper /etc/nova/nova.conf in order to be able to access the database.

openstack-mkpasswd

Hash a given password and output it in a format suitable for direct insertion in Keystone's user database.

This is primarily meant to be used in all cases where users cannot change the password themselves: with this little utility, an OpenStack admin can replace (or set anew) a password entry without being told the cleartext in any way.

openstack-tools's Issues

Take account of users

Hello

Thanks for making this tool available.

One problem I found when using it today was that it did not take account of different users within a tenant, meaning that when I used the 'update' parameter, the usage it entered into the database was much higher than the real one.

I'm using Icehouse, in case that makes a difference.

Adam