OpenStack Operations Guide (2014)

Part II. Operations

Chapter 10. User-Facing Operations

This guide is for OpenStack operators and does not seek to be an exhaustive reference for users, but as an operator, you should have a basic understanding of how to use the cloud facilities. This chapter looks at OpenStack from a basic user perspective, which helps you understand your users’ needs and determine, when you get a trouble ticket, whether it is a user issue or a service issue. The main concepts covered are images, flavors, security groups, block storage, and instances.

Images

OpenStack images can often be thought of as “virtual machine templates.” Images can also be standard installation media such as ISO images. Essentially, they contain bootable file systems that are used to launch instances.

Adding Images

Several premade images exist and can easily be imported into the Image Service. A common image to add is the CirrOS image, which is very small and used for testing purposes. To add this image, simply do:

$ wget http://download.cirros-cloud.net/0.3.1/cirros-0.3.1-x86_64-disk.img

$ glance image-create --name='cirros image' --is-public=true \

  --container-format=bare --disk-format=qcow2 < cirros-0.3.1-x86_64-disk.img

The glance image-create command provides a large set of options for working with your image. For example, the min-disk option is useful for images that require root disks of a certain size (for example, large Windows images). To view these options, do:

$ glance help image-create

The location option is important to note. It does not copy the entire image into the Image Service, but references an original location where the image can be found. Upon launching an instance of that image, the Image Service accesses the image from the location specified.

The copy-from option copies the image from the location specified into the /var/lib/glance/images directory. The same thing is done when using the STDIN redirection with <, as shown in the example.

Run the following command to view the properties of existing images:

$ glance details

Sharing Images Between Projects

In a multitenant cloud environment, users sometimes want to share their personal images or snapshots with other projects. This can be done on the command line with the glance tool by the owner of the image.

To share an image or snapshot with another project, do the following:

1.    Obtain the UUID of the image:

$ glance image-list

2.    Obtain the UUID of the project with which you want to share your image. Unfortunately, nonadmin users are unable to use the keystone command to do this. The easiest solution is to obtain the UUID either from an administrator of the cloud or from a user located in the project.

3.    Once you have both pieces of information, run the glance command:

$ glance member-create <image-uuid> <project-uuid>

For example:

$ glance member-create 733d1c44-a2ea-414b-aca7-69decf20d810 \

    771ed149ef7e4b2b88665cc1c98f77ca

Project 771ed149ef7e4b2b88665cc1c98f77ca will now have access to image 733d1c44-a2ea-414b-aca7-69decf20d810.

Deleting Images

To delete an image, just execute:

$ glance image-delete <image uuid>

NOTE

Deleting an image does not affect instances or snapshots that were based on the image.

Other CLI Options

A full set of options can be found using:

$ glance help

or the OpenStack Image Service CLI Guide.

The Image Service and the Database

The only thing that the Image Service does not store in a database is the image itself. The Image Service database has two main tables:

§  images

§  image_properties

Working directly with the database and SQL queries can provide you with custom lists and reports of images. Technically, you can update properties about images through the database, although this is not generally recommended.

Example Image Service Database Queries

One interesting example is modifying the table of images and the owner of that image. This can be easily done if you simply display the unique ID of the owner. This example goes one step further and displays the readable name of the owner:

mysql> select glance.images.id,

              glance.images.name, keystone.tenant.name, is_public from

              glance.images inner join keystone.tenant on

              glance.images.owner=keystone.tenant.id;

Another example is displaying all properties for a certain image:

mysql> select name, value from

              image_properties where id = <image_id>

Flavors

Virtual hardware templates are called “flavors” in OpenStack, defining sizes for RAM, disk, number of cores, and so on. The default install provides five flavors.

These are configurable by admin users (the rights may also be delegated to other users by redefining the access controls for compute_extension:flavormanage in /etc/nova/policy.json on the nova-api server). To get the list of available flavors on your system, run:

$ nova flavor-list

+----+-----------+-----------+------+-----------+\+-------+-\+-------------+

| ID | Name      | Memory_MB | Disk | Ephemeral |/| VCPUs | /| extra_specs |

+----+-----------+-----------+------+-----------+\+-------+-\+-------------+

| 1  | m1.tiny   | 512       | 1    | 0         |/| 1     | /| {}          |

| 2  | m1.small  | 2048      | 10   | 20        |\| 1     | \| {}          |

| 3  | m1.medium | 4096      | 10   | 40        |/| 2     | /| {}          |

| 4  | m1.large  | 8192      | 10   | 80        |\| 4     | \| {}          |

| 5  | m1.xlarge | 16384     | 10   | 160       |/| 8     | /| {}          |

+----+-----------+-----------+------+-----------+\+-------+-\+-------------+

The nova flavor-create command allows authorized users to create new flavors. Additional flavor manipulation commands can be shown with the command:

$ nova help | grep flavor

Flavors define a number of parameters, resulting in the user having a choice of what type of virtual machine to run—just like they would have if they were purchasing a physical server. Table 10-1 lists the elements that can be set. Note in particular extra_specs, which can be used to define free-form characteristics, giving a lot of flexibility beyond just the size of RAM, CPU, and Disk.

Column

Description

ID

A unique numeric ID.

Name

A descriptive name, such as xx.size_name, is conventional but not required, though some third-party tools may rely on it.

Memory_MB

Virtual machine memory in megabytes.

Disk

Virtual root disk size in gigabytes. This is an ephemeral disk the base image is copied into. You don’t use it when you boot from a persistent volume. The “0” size is a special case that uses the native base image size as the size of the ephemeral root volume.

Ephemeral

Specifies the size of a secondary ephemeral data disk. This is an empty, unformatted disk and exists only for the life of the instance.

Swap

Optional swap space allocation for the instance.

VCPUs

Number of virtual CPUs presented to the instance.

RXTX_Factor

Optional property that allows created servers to have a different bandwidth cap from that defined in the network they are attached to. This factor is multiplied by the rxtx_base property of the network. Default value is 1.0 (that is, the same as the attached network).

Is_Public

Boolean value that indicates whether the flavor is available to all users or private. Private flavors do not get the current tenant assigned to them. Defaults to True.

extra_specs

Additional optional restrictions on which compute nodes the flavor can run on. This is implemented as key-value pairs that must match against the corresponding key-value pairs on compute nodes. Can be used to implement things like special resources (such as flavors that can run only on compute nodes with GPU hardware).

Table 10-1. Flavor parameters

Private Flavors

A user might need a custom flavor that is uniquely tuned for a project she is working on. For example, the user might require 128 GB of memory. If you create a new flavor as described above, the user would have access to the custom flavor, but so would all other tenants in your cloud. Sometimes this sharing isn’t desirable. In this scenario, allowing all users to have access to a flavor with 128 GB of memory might cause your cloud to reach full capacity very quickly. To prevent this, you can restrict access to the custom flavor using the nova command:

$ nova flavor-access-add <flavor-id> <project-id>

To view a flavor’s access list, do the following:

$ nova flavor-access-list <flavor-id>

Best Practices

Once access to a flavor has been restricted, no other projects besides the ones granted explicit access will be able to see the flavor. This includes the admin project. Make sure to add the admin project in addition to the original project.

It’s also helpful to allocate a specific numeric range for custom and private flavors. On UNIX-based systems, nonsystem accounts usually have a UID starting at 500. A similar approach can be taken with custom flavors. This helps you easily identify which flavors are custom, private, and public for the entire cloud.

How Do I Modify an Existing Flavor?

The OpenStack dashboard simulates the ability to modify a flavor by deleting an existing flavor and creating a new one with the same name.

Security Groups

A common new-user issue with OpenStack is failing to set an appropriate security group when launching an instance. As a result, the user is unable to contact the instance on the network.

Security groups are sets of IP filter rules that are applied to an instance’s networking. They are project specific, and project members can edit the default rules for their group and add new rules sets. All projects have a “default” security group, which is applied to instances that have no other security group defined. Unless changed, this security group denies all incoming traffic.

General Security Groups Configuration

The nova.conf option allow_same_net_traffic (which defaults to true) globally controls whether the rules apply to hosts that share a network. When set to true, hosts on the same subnet are not filtered and are allowed to pass all types of traffic between them. On a flat network, this allows all instances from all projects unfiltered communication. With VLAN networking, this allows access between instances within the same project. If allow_same_net_traffic is set to false, security groups are enforced for all connections. In this case, it is possible for projects to simulate allow_same_net_traffic by configuring their default security group to allow all traffic from their subnet.

TIP

As noted in the previous chapter, the number of rules per security group is controlled by the quota_security_group_rules, and the number of allowed security groups per project is controlled by the quota_security_groups quota.

End-User Configuration of Security Groups

Security groups for the current project can be found on the OpenStack dashboard under Access & Security. To see details of an existing group, select the edit action for that security group. Obviously, modifying existing groups can be done from this edit interface. There is a Create Security Group button on the main Access & Security page for creating new groups. We discuss the terms used in these fields when we explain the command-line equivalents.

From the command line, you can get a list of security groups for the project you’re acting in using the nova command:

$ nova secgroup-list

+---------+-------------+

| Name    | Description |

+---------+-------------+

| default | default     |

| open    | all ports   |

+---------+-------------+

To view the details of the “open” security group:

$ nova secgroup-list-rules open

+-------------+-----------+---------+-----------+--------------+

| IP Protocol | From Port | To Port | IP Range  | Source Group |

+-------------+-----------+---------+-----------+--------------+

| icmp        | -1        | 255     | 0.0.0.0/0 |              |

| tcp         | 1         | 65535   | 0.0.0.0/0 |              |

| udp         | 1         | 65535   | 0.0.0.0/0 |              |

+-------------+-----------+---------+-----------+--------------+

These rules are all “allow” type rules, as the default is deny. The first column is the IP protocol (one of icmp, tcp, or udp), and the second and third columns specify the affected port range. The fourth column specifies the IP range in CIDR format. This example shows the full port range for all protocols allowed from all IPs.

When adding a new security group, you should pick a descriptive but brief name. This name shows up in brief descriptions of the instances that use it where the longer description field often does not. Seeing that an instance is using security group http is much easier to understand thanbobs_group or secgrp1.

As an example, let’s create a security group that allows web traffic anywhere on the Internet. We’ll call this group global_http, which is clear and reasonably concise, encapsulating what is allowed and from where. From the command line, do:

$ nova secgroup-create \

       global_http "allow web traffic from the Internet"

+-------------+-------------------------------------+

| Name        | Description                         |

+-------------+-------------------------------------+

| global_http | allow web traffic from the Internet |

+-------------+-------------------------------------+

This creates the empty security group. To make it do what we want, we need to add some rules:

$ nova secgroup-add-rule <secgroup> <ip-proto> <from-port> <to-port> <cidr>

$ nova secgroup-add-rule global_http tcp 80 80 0.0.0.0/0

+-------------+-----------+---------+-----------+--------------+

| IP Protocol | From Port | To Port | IP Range  | Source Group |

+-------------+-----------+---------+-----------+--------------+

| tcp         | 80        | 80      | 0.0.0.0/0 |              |

+-------------+-----------+---------+-----------+--------------+

Note that the arguments are positional, and the from-port and to-port arguments specify the allowed local port range connections. These arguments are not indicating source and destination ports of the connection. More complex rule sets can be built up through multiple invocations ofnova secgroup-add-rule. For example, if you want to pass both http and https traffic, do this:

$ nova secgroup-add-rule global_http tcp 443 443 0.0.0.0/0

+-------------+-----------+---------+-----------+--------------+

| IP Protocol | From Port | To Port | IP Range  | Source Group |

+-------------+-----------+---------+-----------+--------------+

| tcp         | 443       | 443     | 0.0.0.0/0 |              |

+-------------+-----------+---------+-----------+--------------+

Despite only outputting the newly added rule, this operation is additive:

$ nova secgroup-list-rules global_http

+-------------+-----------+---------+-----------+--------------+

| IP Protocol | From Port | To Port | IP Range  | Source Group |

+-------------+-----------+---------+-----------+--------------+

| tcp         | 80        | 80      | 0.0.0.0/0 |              |

| tcp         | 443       | 443     | 0.0.0.0/0 |              |

+-------------+-----------+---------+-----------+--------------+

The inverse operation is called secgroup-delete-rule, using the same format. Whole security groups can be removed with secgroup-delete.

To create security group rules for a cluster of instances, you want to use SourceGroups.

SourceGroups are a special dynamic way of defining the CIDR of allowed sources. The user specifies a SourceGroup (security group name) and then all the users’ other instances using the specified SourceGroup are selected dynamically. This dynamic selection alleviates the need for individual rules to allow each new member of the cluster.

The code is structured like this: nova secgroup-add-group-rule <secgroup> <source-group> <ip-proto> <from-port> <to-port>. An example usage is shown here:

$ nova secgroup-add-group-rule cluster global-http tcp 22 22

The “cluster” rule allows SSH access from any other instance that uses the global-http group.

Block Storage

OpenStack volumes are persistent block-storage devices that may be attached and detached from instances, but they can be attached to only one instance at a time. Similar to an external hard drive, they do not provide shared storage in the way a network file system or object store does. It is left to the operating system in the instance to put a file system on the block device and mount it, or not.

As with other removable disk technology, it is important that the operating system is not trying to make use of the disk before removing it. On Linux instances, this typically involves unmounting any file systems mounted from the volume. The OpenStack volume service cannot tell whether it is safe to remove volumes from an instance, so it does what it is told. If a user tells the volume service to detach a volume from an instance while it is being written to, you can expect some level of file system corruption as well as faults from whatever process within the instance was using the device.

There is nothing OpenStack-specific in being aware of the steps needed to access block devices from within the instance operating system, potentially formatting them for first use and being cautious when removing them. What is specific is how to create new volumes and attach and detach them from instances. These operations can all be done from the Volumes page of the dashboard or by using the cinder command-line client.

To add new volumes, you need only a name and a volume size in gigabytes. Either put these into the create volume web form or use the command line:

$ cinder create --display-name test-volume 10

This creates a 10 GB volume named test-volume. To list existing volumes and the instances they are connected to, if any:

$ cinder list

+------------+---------+--------------------+------+-------------+-------------+

|     ID     | Status  |    Display Name    | Size | Volume Type | Attached to |

+------------+---------+--------------------+------+-------------+-------------+

| 0821...19f |  active |    test-volume     |  10  |     None    |             |

+------------+---------+--------------------+------+-------------+-------------+

OpenStack Block Storage also allows for creating snapshots of volumes. Remember that this is a block-level snapshot that is crash consistent, so it is best if the volume is not connected to an instance when the snapshot is taken and second best if the volume is not in use on the instance it is attached to. If the volume is under heavy use, the snapshot may have an inconsistent file system. In fact, by default, the volume service does not take a snapshot of a volume that is attached to an image, though it can be forced to. To take a volume snapshot, either select Create Snapshot from the actions column next to the volume name on the dashboard volume page, or run this from the command line:

usage: cinder snapshot-create [--force <True|False>]

[--display-name <display-name>]

[--display-description <display-description>]

<volume-id>

Add a new snapshot.

Positional arguments:  <volume-id>         ID of the volume to snapshot

Optional arguments:  --force <True|False>  Optional flag to indicate whether to

                                           snapshot a volume even if its

                                           attached to an instance.

                                           (Default=False) 

--display-name <display-name>              Optional snapshot name.

                                           (Default=None)

--display-description <display-description>

Optional snapshot description. (Default=None)

Block Storage Creation Failures

If a user tries to create a volume and the volume immediately goes into an error state, the best way to troubleshoot is to grep the cinder log files for the volume’s UUID. First try the log files on the cloud controller, and then try the storage node where the volume was attempted to be created:

# grep  903b85d0-bacc-4855-a261-10843fc2d65b /var/log/cinder/*.log

Instances

Instances are the running virtual machines within an OpenStack cloud. This section deals with how to work with them and their underlying images, their network properties, and how they are represented in the database.

Starting Instances

To launch an instance, you need to select an image, a flavor, and a name. The name needn’t be unique, but your life will be simpler if it is because many tools will use the name in place of the UUID so long as the name is unique. You can start an instance from the dashboard from the Launch Instance button on the Instances page or by selecting the Launch action next to an image or snapshot on the Images & Snapshots page.

On the command line, do this:

$ nova boot --flavor <flavor> --image <image> <name>

There are a number of optional items that can be specified. You should read the rest of this section before trying to start an instance, but this is the base command that later details are layered upon.

To delete instances from the dashboard, select the Terminate instance action next to the instance on the Instances page. From the command line, do this:

$ nova delete <instance-uuid>

It is important to note that powering off an instance does not terminate it in the OpenStack sense.

Instance Boot Failures

If an instance fails to start and immediately moves to an error state, there are a few different ways to track down what has gone wrong. Some of these can be done with normal user access, while others require access to your log server or compute nodes.

The simplest reasons for nodes to fail to launch are quota violations or the scheduler being unable to find a suitable compute node on which to run the instance. In these cases, the error is apparent when you run a nova show on the faulted instance:

$ nova show test-instance

+------------------------+-----------------------------------------------------\

| Property               | Value                                               /

+------------------------+-----------------------------------------------------\

| OS-DCF:diskConfig      | MANUAL                                              /

| OS-EXT-STS:power_state | 0                                                   \

| OS-EXT-STS:task_state  | None                                                /

| OS-EXT-STS:vm_state    | error                                               \

| accessIPv4             |                                                     /

| accessIPv6             |                                                     \

| config_drive           |                                                     /

| created                | 2013-03-01T19:28:24Z                                \

| fault                  | {u'message': u'NoValidHost', u'code': 500, u'created/

| flavor                 | xxl.super (11)                                      \

| hostId                 |                                                     /

| id                     | 940f3b2f-bd74-45ad-bee7-eb0a7318aa84                \

| image                  | quantal-test (65b4f432-7375-42b6-a9b8-7f654a1e676e) /

| key_name               | None                                                \

| metadata               | {}                                                  /

| name                   | test-instance                                       \

| security_groups        | [{u'name': u'default'}]                             /

| status                 | ERROR                                               \

| tenant_id              | 98333a1a28e746fa8c629c83a818ad57                    /

| updated                | 2013-03-01T19:28:26Z                                \

| user_id                | a1ef823458d24a68955fec6f3d390019                    /

+------------------------+-----------------------------------------------------\

In this case, looking at the fault message shows NoValidHost, indicating that the scheduler was unable to match the instance requirements.

If nova show does not sufficiently explain the failure, searching for the instance UUID in the nova-compute.log on the compute node it was scheduled on or the nova-scheduler.log on your scheduler hosts is a good place to start looking for lower-level problems.

Using nova show as an admin user will show the compute node the instance was scheduled on as hostId. If the instance failed during scheduling, this field is blank.

Using Instance-Specific Data

There are two main types of instance-specific data: metadata and user data.

Instance metadata

For Compute, instance metadata is a collection of key-value pairs associated with an instance. Compute reads and writes to these key-value pairs any time during the instance lifetime, from inside and outside the instance, when the end user uses the Compute API to do so. However, you cannot query the instance-associated key-value pairs with the metadata service that is compatible with the Amazon EC2 metadata service.

For an example of instance metadata, users can generate and register SSH keys using the nova command:

$ nova keypair-add mykey > mykey.pem

This creates a key named mykey, which you can associate with instances. The file mykey.pem is the private key, which should be saved to a secure location because it allows root access to instances the mykey key is associated with.

Use this command to register an existing key with OpenStack:

$ nova keypair-add --pub-key mykey.pub mykey

NOTE

You must have the matching private key to access instances associated with this key.

To associate a key with an instance on boot, add --key_name mykey to your command line. For example:

$ nova boot --image ubuntu-cloudimage --flavor 2 --key_name mykey myimage

When booting a server, you can also add arbitrary metadata so that you can more easily identify it among other running instances. Use the --meta option with a key-value pair, where you can make up the string for both the key and the value. For example, you could add a description and also the creator of the server:

$ nova boot --image=test-image --flavor=1 \

  --meta description='Small test image' smallimage

When viewing the server information, you can see the metadata included on the metadata line:

$ nova show smallimage

+------------------------+-----------------------------------------+

|     Property           |                   Value                 |

+------------------------+-----------------------------------------+

|   OS-DCF:diskConfig    |               MANUAL                    |

| OS-EXT-STS:power_state |                 1                       |

| OS-EXT-STS:task_state  |                None                     |

|  OS-EXT-STS:vm_state   |               active                    |

|    accessIPv4          |                                         |

|    accessIPv6          |                                         |

|      config_drive      |                                         |

|     created            |            2012-05-16T20:48:23Z         |

|      flavor            |              m1.small                   |

|      hostId            |             de0...487                   |

|        id              |             8ec...f915                  |

|      image             |             natty-image                 |

|     key_name           |                                         |

|     metadata           | {u'description': u'Small test image'}   |

|       name             |             smallimage                  |

|    private network     |            172.16.101.11                |

|     progress           |                 0                       |

|     public network     |             10.4.113.11                 |

|      status            |               ACTIVE                    |

|    tenant_id           |             e83...482                   |

|     updated            |            2012-05-16T20:48:35Z         |

|     user_id            |          de3...0a9                      |

+------------------------+-----------------------------------------+

Instance user data

The user-data key is a special key in the metadata service that holds a file that cloud-aware applications within the guest instance can access. For example, cloudinit is an open source package from Ubuntu, but available in most distributions, that handles early initialization of a cloud instance that makes use of this user data.

This user data can be put in a file on your local system and then passed in at instance creation with the flag --user-data <user-data-file>. For example:

$ nova boot --image ubuntu-cloudimage --flavor 1 --user-data mydata.file

To understand the difference between user data and metadata, realize that user data is created before an instance is started. User data is accessible from within the instance when it is running. User data can be used to store configuration, a script, or anything the tenant wants.

File injection

Arbitrary local files can also be placed into the instance file system at creation time by using the --file <dst-path=src-path> option. You may store up to five files.

For example, let’s say you have a special authorized_keys file named special_authorized_keysfile that for some reason you want to put on the instance instead of using the regular SSH key injection. In this case, you can use the following command:

$ nova boot --image ubuntu-cloudimage --flavor 1  \

  --file /root/.ssh/authorized_keys=special_authorized_keysfile

Associating Security Groups

Security groups, as discussed earlier, are typically required to allow network traffic to an instance, unless the default security group for a project has been modified to be more permissive.

Adding security groups is typically done on instance boot. When launching from the dashboard, you do this on the Access & Security tab of the Launch Instance dialog. When launching from the command line, append --security-groups with a comma-separated list of security groups.

It is also possible to add and remove security groups when an instance is running. Currently this is only available through the command-line tools. Here is an example:

$ nova add-secgroup <server> <securitygroup>

$ nova remove-secgroup <server> <securitygroup>

Floating IPs

Where floating IPs are configured in a deployment, each project will have a limited number of floating IPs controlled by a quota. However, these need to be allocated to the project from the central pool prior to their use—usually by the administrator of the project. To allocate a floating IP to a project, use the Allocate IP to Project button on the Access & Security page of the dashboard. The command line can also be used:

$ nova floating-ip-create

Once allocated, a floating IP can be assigned to running instances from the dashboard either by selecting Associate Floating IP from the actions drop-down next to the IP on the Access & Security page or by making this selection next to the instance you want to associate it with on theInstances page. The inverse action, Dissociate Floating IP, is available only from the Access & Security page and not from the Instances page.

To associate or disassociate a floating IP with a server from the command line, use the following commands:

$ nova add-floating-ip <server> <address>

$ nova remove-floating-ip <server> <address>

Attaching Block Storage

You can attach block storage to instances from the dashboard on the Volumes page. Click the Edit Attachments action next to the volume you want to attach.

To perform this action from command line, run the following command:

$ nova volume-attach <server> <volume> <device>

You can also specify block device mapping at instance boot time through the nova command-line client with this option set:

--block-device-mapping <dev-name=mapping>

The block device mapping format is <dev-name>=<id>:<type>:<size(GB)>:<delete-on-terminate>, where:

dev-name

A device name where the volume is attached in the system at /dev/dev_name

id

The ID of the volume to boot from, as shown in the output of nova volume-list

type

Either snap, which means that the volume was created from a snapshot, or anything other than snap (a blank string is valid). In the preceding example, the volume was not created from a snapshot, so we leave this field blank in our following example.

size (GB)

The size of the volume in gigabytes. It is safe to leave this blank and have the Compute Service infer the size.

delete-on-terminate

A boolean to indicate whether the volume should be deleted when the instance is terminated. True can be specified as True or 1. False can be specified as False or 0.

The following command will boot a new instance and attach a volume at the same time. The volume of ID 13 will be attached as /dev/vdc. It is not a snapshot, does not specify a size, and will not be deleted when the instance is terminated:

$ nova boot --image 4042220e-4f5e-4398-9054-39fbd75a5dd7 \

            --flavor 2 --key-name mykey --block-device-mapping vdc=13:::0 \

            boot-with-vol-test

If you have previously prepared block storage with a bootable file system image, it is even possible to boot from persistent block storage. The following command boots an image from the specified volume. It is similar to the previous command, but the image is omitted and the volume is now attached as /dev/vda:

$ nova boot --flavor 2 --key-name mykey \

            --block-device-mapping vda=13:::0 boot-from-vol-test

Read more detailed instructions for launching an instance from a bootable volume in the OpenStack End User Guide.

To boot normally from an image and attach block storage, map to a device other than vda. You can find instructions for launching an instance and attaching a volume to the instance and for copying the image to the attached volume in the OpenStack End User Guide.

Taking Snapshots

The OpenStack snapshot mechanism allows you to create new images from running instances. This is very convenient for upgrading base images or for taking a published image and customizing it for local use. To snapshot a running instance to an image using the CLI, do this:

$ nova image-create <instance name or uuid> <name of new image>

The dashboard interface for snapshots can be confusing because the Images & Snapshots page splits content up into several areas:

§  Images

§  Instance snapshots

§  Volume snapshots

However, an instance snapshot is an image. The only difference between an image that you upload directly to the Image Service and an image that you create by snapshot is that an image created by snapshot has additional properties in the glance database. These properties are found in theimage_properties table and include:

Name

Value

image_type

snapshot

instance_uuid

<uuid of instance that was snapshotted>

base_image_ref

<uuid of original image of instance that was snapshotted>

image_location

snapshot

Live Snapshots

Live snapshots is a feature that allows users to snapshot the running virtual machines without pausing them. These snapshots are simply disk-only snapshots. Snapshotting an instance can now be performed with no downtime (assuming QEMU 1.3+ and libvirt 1.0+ are used).

ENSURING SNAPSHOTS ARE CONSISTENT

The following section is from Sébastien Han’s “OpenStack: Perform Consistent Snapshots” blog entry.

A snapshot captures the state of the file system, but not the state of the memory. Therefore, to ensure your snapshot contains the data that you want, before your snapshot you need to ensure that:

§  Running programs have written their contents to disk

§  The file system does not have any “dirty” buffers: where programs have issued the command to write to disk, but the operating system has not yet done the write

To ensure that important services have written their contents to disk (such as databases), we recommend that you read the documentation for those applications to determine what commands to issue to have them sync their contents to disk. If you are unsure how to do this, the safest approach is to simply stop these running services normally.

To deal with the “dirty” buffer issue, we recommend using the sync command before snapshotting:

# sync

Running sync writes dirty buffers (buffered blocks that have been modified but not written yet to the disk block) to disk.

Just running sync is not enough to ensure that the file system is consistent. We recommend that you use the fsfreeze tool, which halts new access to the file system, and create a stable image on disk that is suitable for snapshotting. The fsfreeze tool supports several file systems, including ext3, ext4, and XFS. If your virtual machine instance is running on Ubuntu, install the util-linux package to get fsfreeze:

# apt-get install util-linux

If your operating system doesn’t have a version of fsfreeze available, you can use xfs_freeze instead, which is available on Ubuntu in the xfsprogs package. Despite the “xfs” in the name, xfs_freeze also works on ext3 and ext4 if you are using a Linux kernel version 2.6.29 or greater, since it works at the virtual file system (VFS) level starting at 2.6.29. The xfs_freeze version supports the same command-line arguments as fsfreeze.

Consider the example where you want to take a snapshot of a persistent block storage volume, detected by the guest operating system as /dev/vdb and mounted on /mnt. The fsfreeze command accepts two arguments:

-f

Freeze the system

-u

Thaw (unfreeze) the system

To freeze the volume in preparation for snapshotting, you would do the following, as root, inside the instance:

# fsfreeze -f /mnt

You must mount the file system before you run the fsfreeze command.

When the fsfreeze -f command is issued, all ongoing transactions in the file system are allowed to complete, new write system calls are halted, and other calls that modify the file system are halted. Most importantly, all dirty data, metadata, and log information are written to disk.

Once the volume has been frozen, do not attempt to read from or write to the volume, as these operations hang. The operating system stops every I/O operation and any I/O attempts are delayed until the file system has been unfrozen.

Once you have issued the fsfreeze command, it is safe to perform the snapshot. For example, if your instance was named mon-instance and you wanted to snapshot it to an image named mon-snapshot, you could now run the following:

$ nova image-create mon-instance mon-snapshot

When the snapshot is done, you can thaw the file system with the following command, as root, inside of the instance:

# fsfreeze -u /mnt

If you want to back up the root file system, you can’t simply run the preceding command because it will freeze the prompt. Instead, run the following one-liner, as root, inside the instance:

# fsfreeze -f / && sleep 30 && fsfreeze -u /

Instances in the Database

While instance information is stored in a number of database tables, the table you most likely need to look at in relation to user instances is the instances table.

The instances table carries most of the information related to both running and deleted instances. It has a bewildering array of fields; for an exhaustive list, look at the database. These are the most useful fields for operators looking to form queries:

§  The deleted field is set to 1 if the instance has been deleted and NULL if it has not been deleted. This field is important for excluding deleted instances from your queries.

§  The uuid field is the UUID of the instance and is used throughout other tables in the database as a foreign key. This ID is also reported in logs, the dashboard, and command-line tools to uniquely identify an instance.

§  A collection of foreign keys are available to find relations to the instance. The most useful of these—user_id and project_id—are the UUIDs of the user who launched the instance and the project it was launched in.

§  The host field tells which compute node is hosting the instance.

§  The hostname field holds the name of the instance when it is launched. The display-name is initially the same as hostname but can be reset using the nova rename command.

A number of time-related fields are useful for tracking when state changes happened on an instance:

§  created_at

§  updated_at

§  deleted_at

§  scheduled_at

§  launched_at

§  terminated_at

Good Luck!

This section was intended as a brief introduction to some of the most useful of many OpenStack commands. For an exhaustive list, please refer to the Admin User Guide, and for additional hints and tips, see the Cloud Admin Guide. We hope your users remain happy and recognize your hard work! (For more hard work, turn the page to the next chapter, where we discuss the system-facing operations: maintenance, failures and debugging.)