struct HealthCheckRecord

Defined at line 615 of file ../../src/devices/bin/driver_manager/node.h

The latest health check outcome for a node. driver_manager caches it so that failures such

as a wedged driver dispatcher or a crashed driver host are preserved out-of-process.

Lifetime: the record belongs to one driver instance. It is cleared, along with the alert

probe timestamps, whenever a new driver is started on the node (first bind, rebind or

restart), so a stale TIMEOUT or UNHEALTHY from the previous instance never outlives it.

Replacement policy (see MaybeStoreHealthCheckRecord):

(a) A result that carries no health information (NOT_SUPPORTED, UNAVAILABLE,

COMMUNICATION_ERROR, INVALID_STATE, NOT_FOUND) never replaces a record that does carry

some (a status, or a synthetic TIMEOUT / host-closed / transport-failure condition). It is

only cached while there is no informative record yet.

(b) An OK PING result does not hide a non-OK record from a FAST or DEEP check or a

self_fault alert: a liveness ping says nothing about the hardware. The most recent

informative non-PING record is retained separately for this purpose, so a PING failure

(TIMEOUT, crashed host) in between does not launder the hardware fault away: once the

PING succeeds again the retained non-PING record is published again.

All other results replace the existing record.

Public Members

optional status

optional error

optional level

basic_time timestamp

vector diagnostics

vector conditions

Public Methods

bool is_ping ()

True if the record came from a PING, i.e. it only measures dispatcher liveness.

Defined at line 641 of file ../../src/devices/bin/driver_manager/node.h

HealthCheckRecord FromResponse (const fuchsia_driver_framework::HealthCheckResponse & response, std::optional<fuchsia_driver_framework::HealthCheckRequest> level, zx::time timestamp)

Builds a record from a response returned by (or synthesized for) the driver. |level| is the

level of the check that produced it, or nullopt for a SELF_FAULT alert.

Defined at line 3548 of file ../../src/devices/bin/driver_manager/node.cc

HealthCheckRecord FromError (fuchsia_driver_framework::HealthCheckError error, std::optional<fuchsia_driver_framework::HealthCheckRequest> level, zx::time timestamp)

Builds a record for a check that failed with |error|. TIMEOUT and other errors that

indicate a problem add a single synthetic UNHEALTHY condition; NOT_SUPPORTED, UNAVAILABLE,

COMMUNICATION_ERROR, INVALID_STATE and NOT_FOUND carry no health information and add none.

Defined at line 3562 of file ../../src/devices/bin/driver_manager/node.cc

fuchsia_driver_framework::HealthCondition SyntheticCondition (std::string reason_code, std::string message)

Builds an UNHEALTHY condition that driver_manager synthesizes for failures it detects

itself (e.g. a crashed driver host), as opposed to conditions reported by the driver.

Defined at line 3505 of file ../../src/devices/bin/driver_manager/node.cc

std::string_view first_reason_code ()

Returns the reason code of the first condition, or "-" if there is none. Used for logs.

Defined at line 3514 of file ../../src/devices/bin/driver_manager/node.cc

bool is_informative ()

True if the record says something about the driver's health: it has a status, or a

synthetic condition for a failure driver_manager detected. False for the bare errors listed

under replacement policy (a).

Defined at line 3521 of file ../../src/devices/bin/driver_manager/node.cc