salutem.core — io.logicblocks/salutem.core 0.1.9-RC2

A system for defining and maintaining a collection of health checks with support for both realtime and background checks.

The salutem.core namespace is the public interface of the system and should be used in preference of the contained namespaces.

salutem is somewhat inspired by dropwizard-health which may provide additional insight into its design.

all-checks clj

(all-checks registry)

Returns the set of checks present in the registry.

background-check clj

(background-check check-name check-fn)

(background-check check-name check-fn opts)

Constructs a background check with the provided name and check function.

A background check is one that is evaluated periodically with the result cached in a registry until the next evaluation, conducted by a maintenance pipeline, which will occur once the time to re-evaluation of the check has passed.

Background checks are useful for external dependencies where it is important not to perform the check too frequently and where the health status only needs to be accurate on the order of the time to re-evaluation.

Takes the following parameters:

• check-name: a keyword representing the name of the check
• check-fn: an arity-2 function, with the first argument being a context map as provided during evaluation or at maintenance pipeline construction and the second argument being a callback function which should be called with the result of the check to signal the check is complete; note, check functions must be non-blocking.
• opts: an optional map of additional options for the check, containing:
  • :salutem/timeout: a duration representing the amount of time to wait for the check to complete before considering it failed, defaulting to 10 seconds.
  • :salutem/time-to-re-evaluation: a duration representing the time to wait after a check is evaluated before attempting to re-evaluate it, defaulting to 10 seconds.

Any extra entries provided in the opts map are retained on the check for later use.

Note that a result for a background check may live for longer than the time to re-evaluation since evaluation takes time and the result will continue to be returned from the registry whenever the check is resolved until the evaluation has completed and the new result has been added to the registry.

background? clj

(background? check)

Returns true if the provided check is a background check, false otherwise.

check-name clj

(check-name check)

Returns the name of the provided check.

check-names clj

(check-names registry)

Returns the set of check names present in the registry.

duration clj

(duration amount unit)

Constructs an object representing a duration of time.

This object is used to specify, for example, the time to re-evaluation and timeout on a check or the interval passed to a maintenance pipeline.

Takes an amount and a unit:

• amount is the length of the duration, measured in terms of the unit
• unit is one of :nanos, :micros, :millis, :seconds, :minutes, :hours, :half-days, :days, :weeks, :months, :years, :decades, :centuries, :millennia, :eras or :forever

Note: internally, this constructs a java.time.Duration and is merely a convenience function. As such, a java.time.Duration can be passed directly wherever this function would be used.

empty-registry clj

(empty-registry)

Constructs an empty registry which can be populated using with-check and with-cached-result.

evaluate clj

(evaluate check)

(evaluate check context)

(evaluate check context callback-fn)

Evaluates the provided check, returning the result of the evaluation.

Optionally takes a context map containing arbitrary context required by the check in order to run and passed to the check function as the first argument.

By default, the check is evaluated synchronously. If a callback function is provided, the function starts evaluation asynchronously, returns immediately and invokes the callback function with the result once available.

find-cached-result clj

(find-cached-result registry check-name)

Finds the cached result for the check with the given name in the registry. Returns nil if no result can be found or if the check does not exist.

find-check clj

(find-check registry check-name)

Finds the check with the given name in the registry. Returns nil if no check can be found.

healthy clj

(healthy)

(healthy extra-data)

Constructs a healthy result.

The optional map of extra data is stored with the result for future use. Unless overridden in the extra data map, an :salutem/evaluated-at field is added to the result, set to the current date time in the system default time zone.

healthy? clj

(healthy? result)

Returns true if the result has a :healthy status, false otherwise.

maintain clj

(maintain registry-store)

(maintain registry-store options)

Constructs and starts a maintenance pipeline to maintain up-to-date results for the checks in the registry in the provided registry store atom.

The maintenance pipeline consists of a number of independent processes:

• a maintainer which triggers an attempt to refresh the results periodically,
• a refresher which requests evaluation of each outdated check on each refresh attempt,
• an evaluator which evaluates outdated checks to obtain a fresh result,
• an updater which updates the registry store atom with fresh check results,
• a notifier which calls callback functions when fresh check results are available.

The maintenance pipeline can be configured via an optional map which can contain the following options:

• :context: a map containing arbitrary context required by checks in order to run and passed to the check functions as the first argument; defaults to an empty map
• :interval: a duration describing the wait interval between attempts to refresh the results in the registry; defaults to 200 milliseconds
• :notification-callback-fns: a sequence of arity-2 functions, with the first argument being a check and the second argument being a result, which are called whenever a new result is available for a check; empty by default
• :trigger-channel: the channel on which trigger messages are sent, to indicate that a refresh of the registry should be attempted, defaults to a channel with a sliding buffer of length 1, i.e., in the case of a long running attempt, all but the latest trigger message will be dropped from the channel
• :evaluation-channel: the channel on which messages requesting evaluation of checks are sent, defaults to a channel with a buffer of size 10
• :result-channel: the channel on which results are placed after evaluation, defaults to a channel with a buffer of size 10
• :updater-result-channel: a tap of the result-channel which sends result messages on to the updater, defaults to a channel with a buffer of size 10
• :notifier-result-channel: a tap of the result-channel which sends result messages on to the notifier, defaults to a channel with a buffer of size 10

If the context map contains a :logger key with a cartus.core/Logger value, the maintenance pipeline will emit a number of log events throughout operation.

Returns the maintenance pipeline which can be passed to shutdown in order to stop operation.

outdated-checks clj

(outdated-checks registry)

Returns the set of checks that are currently outdated in the registry based on the type of the check and the cached results available.

See outdated? for details on which it means for a check to be outdated.

outdated? clj

(outdated? result check)

(outdated? result check relative-to)

Returns true if the result of the check is outdated, false otherwise.

For a realtime check, a result is always considered outdated.

For a background check, a result is considered outdated if the time to re-evaluation of the check has passed, i.e., if its evaluation date time is before the current date time minus the check's time to re-evaluation.

If relative-to is provided, the calculation is performed relative to that date time rather than to the current date time.

realtime-check clj

(realtime-check check-name check-fn)

(realtime-check check-name check-fn opts)

Constructs a realtime check with the provided name and check function.

A realtime check is one that is re-evaluated whenever the check is resolved, with no caching of results taking place.

Realtime checks are useful when the accuracy of the check needs to be very high or where the check itself is inexpensive.

Takes the following parameters:

• check-name: a keyword representing the name of the check
• check-fn: an arity-2 function, with the first argument being a context map as provided during evaluation or at maintenance pipeline construction and the second argument being a callback function which should be called with the result fo the check to signal the check is complete; note, check functions must be non-blocking.
• opts: an optional map of additional options for the check, containing:
  • :salutem/timeout: a duration representing the amount of time to wait for the check to complete before considering it failed, defaulting to 10 seconds.

Any extra entries provided in the opts map are retained on the check for later use.

realtime? clj

(realtime? check)

Returns true if the provided check is a realtime check, false otherwise.

resolve-check clj

(resolve-check registry check-name)

(resolve-check registry check-name context)

(resolve-check registry check-name context callback-fn)

Resolves a result for the check of the given name in the registry.

If the check is a background check and there is a cached result available, it is returned. If no cached result is available, the check is evaluated in order to obtain a result to return.

If the check is a realtime check, it is always evaluated in order to obtain a result to return and caching is not used.

Optionally takes a context map containing arbitrary context required by the check in order to run and passed to the check function as the first argument.

By default, the check is resolved synchronously. If a callback function is provided, the function starts resolution asynchronously, returns immediately and invokes the callback function with the result once available.

resolve-checks clj

(resolve-checks registry)

(resolve-checks registry context)

(resolve-checks registry context callback-fn)

Resolves all checks in the registry, returning a map of check names to results.

Checks requiring re-evaluation are evaluated in parallel such that this function should take about as long as the slowest check (assuming IO is the dominant blocker).

Optionally takes a context map containing arbitrary context required by checks in order to run and passed to the check functions as the first argument.

By default, the checks are resolved synchronously. If a callback function is provided, the function starts resolution asynchronously, returns immediately and invokes the callback function with the results once available.

See resolve-check for details on how each check is resolved.

result clj

(result status)

(result status extra-data)

Constructs a result with the provided status.

The optional map of extra data is stored with the result for future use. Unless overridden in the extra data map, a :salutem/evaluated-at field is added to the result, set to the current date time in the system default time zone.

shutdown clj

(shutdown maintenance-pipeline)

Shuts down the maintenance pipeline preventing further updates to the registry.

status clj

(status result)

Returns the status of the provided result.

unhealthy clj

(unhealthy)

(unhealthy extra-data)

Constructs an unhealthy result.

The optional map of extra data is stored with the result for future use. Unless overridden in the extra data map, an :salutem/evaluated-at field is added to the result, set to the current date time in the system default time zone.

unhealthy? clj

(unhealthy? result)

Returns true if the result has an :unhealthy status, false otherwise.

with-cached-result clj

(with-cached-result registry check-name result)

Adds the result for the check with the given name to the registry, returning a new registry.

with-check clj

(with-check registry check)

Adds the check to the registry, returning a new registry.