Sensu Frequently Asked Questions

Please note the following frequently asked questions about Sensu Core, Sensu Enterprise, Sensu Training, Professional Services for Sensu, and more. If you need support for Sensu, please consider giving Sensu Enterprise a try.

Do I need RabbitMQ to be installed on every system I wish to monitor?

No. Sensu uses RabbitMQ as a Transport. Sensu services require access to a shared instance of the defined Sensu Transport (e.g. a RabbitMQ cluster) to function. Sensu check requests and check results are published as “messages” to the Sensu Transport, and the corresponding Sensu services receive these messages by subscribing to the appropriate subscription topics.

Does Redis need to be installed on every system I wish to monitor?

No. Sensu uses Redis as a data store, and the Sensu server services (i.e. sensu-server & sensu-api for Sensu Core; sensu-enterprise for Sensu Enterprise) require access to the same Redis instance (or cluster) to store and access the Sensu client registry, check results, check execution history, and current event data.

Do check definitions need to exist on every system I wish to monitor?

No. Check definitions can be written as publish/subscribe (pubsub) or standalone. Pubsub checks, which specify a list of subscribers, need only be configured on the Sensu server. Standalone checks, which are scheduled and executed by the Sensu client, need only be configured on the client(s) where they should be run.

Where should check plugin executables be installed?

Regardless of where checks are defined, the actual check executables need to exist on the filesystem for the Sensu client to execute them. Check plugin executables can be installed in /etc/sensu/plugins or /opt/sensu/embedded/bin, the latter being the location where plugin executables are installed via sensu-install.

What is a standalone check?

A standalone check is a check definition that is installed on and executed by the Sensu client without being scheduled by the Sensu server. Standalone checks defer Check execution scheduling responsibilities to Sensu clients, enabling decentralized management of monitoring checks and distribution of scheduling responsibilities. Standalone checks may be used in conjunction with pubsub checks, and are distinguished from pubsub checks by inclusion of the "standalone": true configuration parameter.

What happens if a single check is defined on both the Sensu server and client?

When a check request is published for a check defined on the Sensu server, the Sensu client will look for a local definition matching the check name prior to executing the check. If a local definition exists, it is merged with the definition provided by the server, with any local definition attributes overriding the definition provided by the Sensu server.

What is Sensu Client safe_mode?

In safe_mode a client will not run a check published by a Sensu server unless that check is also defined on the client. Safe mode must be enabled on the Sensu Client via the safe_mode configuration attribute.

Can multiple Sensu servers be run concurrently, in a cluster?

Yes. Sensu is designed to be scaled horizontally (i.e. by adding additional Sensu servers). It supports fully automated leader election (ensuring that a single Sensu server acts as a centralized Check Execution Scheduler), automated failover (automatically electing a new leader if the previous leader is unexpectedly unavailable), and distributed event processing (check results are distributed across all Sensu servers in a round-robin fashion). Running more than one Sensu server is highly recommended for performance and availability.

How are new systems registered?

Automatically. Sensu clients register themselves when they start up. The Sensu client process requires access to the Sensu Transport (by default, this is RabbitMQ; see Sensu Configuration), and some minimal client configuration (e.g. a name, address, and one or more subscriptions) in order to start. When the Sensu client process starts, it begins sending “keepalives” – a special type of check result containing client configuration data – which the Sensu server uses to know that a client is still connected. When a client keepalive is received for a client name that is not currently registered with Sensu, the client is added to the registry and a registration event is created automatically.

Do system clocks need to be synchronized?

Yes. The Sensu services (i.e. sensu-client, sensu-server, sensu-api, sensu-enterprise) use the local/system clock for generating timestamps. When system clocks are out of sync between Sensu clients (where data is collected) and the Sensu server (where data is processed), Sensu may generate false positive client keepalive events, among other potentially unexpected behaviors. Time synchronization can be facilitated with NTP.

Is Sensu Enterprise available as a hosted / SaaS solution?

No. Like Sensu Core, Sensu Enterprise is installed on your organization’s infrastructure alongside other applications and services. Sensu Enterprise packages are available for major Linux distributions including RHEL, CentOS, Debian and Ubuntu.

Is Sensu available for Microsoft Windows?

Yes. An MSI installer package is available on the Downloads page. Please visit the Sensu documentation for more information on configuring Sensu on Windows.

How do I increase log verbosity?

You can toggle debug logging on and off by sending the Sensu process a TRAP signal.

For example:

$ ps aux | grep [s]ensu-server
sensu     5992  1.7  0.3 177232 24352 ...
$ kill -TRAP 5992

Start your FREE trial today.

Already using Sensu Core? Upgrade to Sensu Enterprise today to take advantage of its enterprise console, added-value features, built-in integrations, FREE annual training, and enterprise-class support.