gcloud beta compute instances ops-agents policies update

gcloud beta compute instances ops-agents policies update - update a Google Cloud operations suite agent (Ops Agent) policy

gcloud beta compute instances ops-agents policies update POLICY_ID [--agent-rules=type=TYPE,version=VERSION,package-state=PACKAGE-STATE,enable-autoupgrade=ENABLE-AUTOUPGRADE;[…]] [--description=DESCRIPTION] [--etag=ETAG] [--os-types=short-name=SHORT-NAME,version=VERSION;[…]] [--clear-group-labels     | --group-labels=[LABEL_NAME=LABEL_VALUE,LABEL_NAME=LABEL_VALUE,…;…]] [--clear-instances     | --instances=[zones/ZONE_NAME/instances/INSTANCE_NAME,…]] [--clear-zones     | --zones=[ZONE_NAME,…]] [GCLOUD_WIDE_FLAG …]

The command returns the content of the updated policy or an error indicating why the update fails. The updated policy takes effect asynchronously. It can take 10-15 minutes for the VMs to enforce the updated policy.

The available flags for the update command are similar to the flags for the create command. All the flags for update are optional. If a flag is not specified, it retains the original value. The full value of each flag needs to be re-stated during update. Take the --agents flag for example:

If the original policy specified two agents (--agents="type=logging;type=metrics"), and only one agent (--agents="type=logging") is specified in a gcloud beta compute instances ops-agents policies update command, then the policy stops managing and enforcing the unspecified agent. In order to remove the metrics agent in this case, set the package state explicitly to removed (--agents="type=logging;type=metrics,package-state=removed").

In order to explicitly clear the --group-labels, --instances, and --zones instance filters, use the following flags as documented below: --clear-group-labels, --clear-instances, and --clear-zones flags.

gcloud beta compute instances ops-agents policies update ops-agents-test-policy --agent-rules="type=logging,enable-autoupgrade=false;type=metrics,enable-autoupgrade=false" --instances=zones/us-central1-a/instances/test-instance --os-types=short-name=centos,version=7

To update a policy named ops-agents-prod-policy to target all CentOS 7 VMs in zone us-central1-a with either env=prod,product=myapp labels or env=staging,product=myapp labels, and make sure the logging agent and metrics agent versions are pinned to specific major versions for staging and production, run:

gcloud beta compute instances ops-agents policies update ops-agents-prod-policy --agent-rules="type=logging,version=1.*.*,enable-autoupgrade=false;type=metrics,version=6.*.*,enable-autoupgrade=false" --group-labels="env=prod,product=myapp;env=staging,product=myapp" --os-types=short-name=centos,version=7 --zones=us-central1-a

To update a policy named ops-agents-labels-policy to clear the instances filters and use a group labels filter instead to target VMs with either env=prod,product=myapp or env=staging,product=myapp labels, run:

gcloud beta compute instances ops-agents policies update ops-agents-labels-policy --clear-instances --group-labels="env=prod,product=myapp;env=staging,product=myapp"

To perform the same update as above, conditionally on the fact that the policy's etag (retrieved by an earlier command) is f59741c8-bb5e-4ee6-bf6f-c4ebeb6b06e0, run:

gcloud beta compute instances ops-agents policies update ops-agents-labels-policy --clear-instances --group-labels="env=prod,product=myapp;env=staging,product=myapp" --etag=f59741c8-bb5e-4ee6-bf6f-c4ebeb6b06e0

POLICY_ID

ID of the policy.

This ID must start with ops-agents-, contain only lowercase letters, numbers, and hyphens, end with a number or a letter, be between 1-63 characters, and be unique within the project. The goal of the prefix ops-agents- is to easily distinguish these Ops Agents specific policies from other generic policies and lower the chance of naming conflicts.

--agent-rules=type=TYPE,version=VERSION,package-state=PACKAGE-STATE,enable-autoupgrade=ENABLE-AUTOUPGRADE;[…]

A non-empty list of agent rules to be enforced by the policy.

This flag must be quoted. Items in the list are separated by ";". Each item in the list is a <key, value> map that represents a logging or metrics agent. The allowed values of the key are as follows.

type

Type of agent to manage.

Required. Allowed values: logging, metrics and ops-agent. Use logging for the Logging Agent (https://cloud.google.com/logging/docs/agent). Use metrics for the Monitoring Agent (https://cloud.google.com/monitoring/agent). Use ops-agent for the Ops Agent (https://cloud.google.com/stackdriver/docs/solutions/ops-agent). The Ops Agent has both a logging module and a metrics module already. So other types of agents are not allowed when there is an agent with type ops-agent. See https://cloud.google.com/stackdriver/docs/solutions/agents#which-agent-should-you-choose for which agent to use.

enable-autoupgrade

Whether to enable autoupgrade of the agent.

Required. Allowed values: true or false. This has to be false if the agent version is set to a specific patch version in the format of version=MAJOR_VERSION.MINOR_VERSION.PATCH_VERSION.

version

Version of the agent to install.

Optional. Default to version=current-major. The allowed values and formats are as follows.

version=latest

With this setting, the latest version of the agent is installed at the time when the policy is applied to an instance.

If multiple instances are created at different times but they all fall into the instance filter rules of an existing policy, they may end up with different versions of the agent, depending on what the latest version of the agent is at the policy application time (in this case the instance creation time). One way to avoid this is to set enable-autoupgrade=true. This guarantees that the installed agents on all instances that are managed by this policy are always up to date and conform to the same version.

While this version=latest setting makes it easier to keep the agent version up to date, this setting does come with a potential risk. When a new major version is released, the policy may install the latest version of the agent from that new major release, which may introduce breaking changes. For production environments, consider using the version=MAJOR_VERSION.*.* setting below for safer agent deployments.

version=MAJOR_VERSION.*.*

With this setting, the latest version of agent from a specific major version is installed at the time when the policy is applied to an instance.

If multiple instances are created at different times but they all fall into the instance filter rules of an existing policy, they may end up with different versions of the agent, depending on what the latest version of the agent is at the policy application time (in this case the instance creation time). One way to avoid this is to set enable-autoupgrade=true. This guarantees that the installed agents on all instances that are managed by this policy are always up to date within that major version and conform to the same version.

When a new major release is out, this setting ensures that only the latest version from the specified major version is installed, which avoids accidentally introducing breaking changes. This is recommended for production environments to ensure safer agent deployments.

version=current-major

With this setting, the version field is automatically set to version=MAJOR_VERSION.*.*, where MAJOR_VERSION is the current latest major version released. Refer to the version=MAJOR_VERSION.*.* section for the expected behavior.

version=MAJOR_VERSION.MINOR_VERSION.PATCH_VERSION

With this setting, the specified exact version of agent is installed at the time when the policy is applied to an instance. enable-autoupgrade must be false for this setting.

This setting is not recommended since it prevents the policy from installing new versions of the agent that include bug fixes and other improvements.

One limitation of this setting is that if the agent gets manually uninstalled from the instances after the policy gets applied, the policy can only ensure that the agent is re-installed. It is not able to restore the expected exact version of the agent.

version=5.5.2-BUILD_NUMBER

Allowed for the metrics agent (type=metrics) only.

With this setting, the specified exact build number of the deprecated 5.5.2 metrics agent is installed at the time when the policy is applied to an instance. enable-autoupgrade must be false for this setting.

This setting is deprecated and will be decommissioned along with the 5.5.2 metrics agent on Apr 28, 2021 (https://cloud.google.com/stackdriver/docs/deprecations/mon-agent). It is not recommended since it prevents the policy from installing new versions of the agent that include bug fixes and other improvements.

One limitation of this setting is that if the agent gets manually uninstalled from the instances after the policy gets applied, the policy can only ensure that the agent is re-installed. It is not able to restore the expected exact version of the agent.

package-state

Desired package state of the agent.

Optional. Default to package-state=installed. The allowed values are as follows.

package-state=installed

With this setting, the policy will ensure the agent package is installed on the instances and the agent service is running.

package-state=removed

With this setting, the policy will ensure the agent package is removed from the instances, which stops the service from running.

--description=DESCRIPTION

Description of the policy.

--etag=ETAG

Etag of the policy.

etag is used for optimistic concurrency control as a way to help prevent simultaneous updates of a policy from overwriting each other. It is strongly suggested that systems make use of the etag in the read-modify-write cycle to perform policy updates in order to avoid race conditions: an etag is returned in the response of a describe command, and systems are expected to put that etag in the request to an update command to ensure that their change will be applied to the same version of the policy.

--os-types=short-name=SHORT-NAME,version=VERSION;[…]

A non-empty list of OS types to filter instances that the policy applies to.

For Alpha and Beta, exactly one OS type needs to be specified. The support for multiple OS types will be added later for more flexibility. Each OS type is defined by the combination of short-name and version fields.

Sample values:

OS Short Name      OS Version
centos             8
centos             7
debian             12
debian             11
debian             10
debian             9
rhel               9.*
rhel               8.*
rhel               7.*
rocky              9.*
rocky              8.*
sles               12.*
sles               15.*
sles_sap           12.*
sles_sap           15.*
ubuntu             16.04
ubuntu             18.04
ubuntu             19.10
ubuntu             20.04
ubuntu             21.04
ubuntu             21.10
ubuntu             22.04
ubuntu             23.04
ubuntu             23.10
ubuntu             24.04
windows            10.*
windows            6.*

short-name

Short name of the OS.

Required. Allowed values: centos, debian, rhel, rocky, sles, sles_sap, ubuntu.

To inspect the exact OS short name of an instance, run:

gcloud beta compute instances os-inventory describe INSTANCE_NAME | grep "^ShortName: "

Under the hood, this value is derived from the ID field in the /etc/os-release file for most operating systems.

version

Version of the OS.

Required. This can be either an exact match or a prefix followed by the * wildcard.

To inspect the exact OS version of an instance, run:

gcloud beta compute instances os-inventory describe INSTANCE_NAME | grep "^Version: "

Under the hood, this value is derived from the VERSION_ID field in the /etc/os-release file for most operating systems.

At most one of these can be specified:

--clear-group-labels

Clear the group labels filter that was previously set by the --group-labels flag to filter instances that the policy applies to.

--group-labels=[LABEL_NAME=LABEL_VALUE,LABEL_NAME=LABEL_VALUE,…;…]

A list of label maps to filter instances that the policy applies to.

Optional. The --group-labels flag needs to be quoted. Each label map item in the list are separated by ;. To manage instance labels, refer to:

gcloud beta compute instances add-labels

gcloud beta compute instances remove-labels

Each label map item in the --group-labels list is a map in the format of LABEL_NAME=LABEL_VALUE,LABEL_NAME=LABEL_VALUE,…. An instance has to match all of the LABEL_NAME=LABEL_VALUE criteria inside a label map to be considered a match for that label map. But the instance only needs to match one label map in the --group-labels list.

For example, --group-labels="env=prod,product=myapp;env=staging,product=myapp" implies the matching criteria is:

(env=prod AND product=myapp) OR (env=staging AND product=myapp)

At most one of these can be specified:

--clear-instances

Clear the instances filter that was previously set by the --instances flag to filter instances that the policy applies to.

--instances=[zones/ZONE_NAME/instances/INSTANCE_NAME,…]

A list of fully-qualified names to filter instances that the policy applies to.

Each item in the list must be in the format of zones/ZONE_NAME/instances/INSTANCE_NAME. The policy can also target instances that are not yet created.

To list all existing instances, run:

gcloud compute instances list

The --instances flag is recommended for use during development and testing. In production environments, it's more common to select instances via a combination of --zones and --group-labels.

At most one of these can be specified:

--clear-zones

Clear the zones filter that was previously set by the --zones flag to filter instances that the policy applies to.

--zones=[ZONE_NAME,…]

A list of zones to filter instances to apply the policy.

To list available zones, run:

gcloud compute zones list

The use of the --zones and --group-labels flags is recommended for production environments. For testing and development, it's more common to select instances directly via the --instances flag.

Run $ gcloud help for details.