CoolCMS API

ElasticPresenter extends BasePresenter
in package

Table of Contents

Properties

$cronLockManager  : CronLockManager
$operationRunner  : OperationRunner
Boundary of one cron action (see {@see self::tryCall()}).
$settingContext  : SettingContext
$settingManager  : SettingManager
$debugMode  : bool
$productionMode  : bool
$advertSource  : AdvertDocumentSource
$esIndexer  : EsEntityIndexer

Methods

__construct()  : mixed
actionReindex()  : void
Full reindex of adverts into a new index version with an atomic alias swap.
actionSync()  : void
Incremental sync: indexes adverts inserted since the watermark and cleans up orphans.
releaseLock()  : void
Releases the lock at most once; called by both `shutdown()` and the PHP shutdown function.
sendResponse()  : void
setConfigParameters()  : void
shutdown()  : void
startup()  : void
isSameSiteRequest()  : bool
Tells whether the request comes from the own site (Nette SameSite cookie); protects GET actions with a side effect.
sendError()  : never
sendSkipped()  : never
Reports that the job did only part of the work and the next call finishes the rest (e.g. a reindex done in chunks).
sendSuccess()  : never
tryCall()  : bool
Runs the `action<Action>()` method as one operation (transitional).

Properties

$operationRunner

Boundary of one cron action (see {@see self::tryCall()}).

public OperationRunner $operationRunner
Attributes
#[Inject]

Methods

actionReindex()

Full reindex of adverts into a new index version with an atomic alias swap.

public actionReindex() : void

Zero downtime; a Redis lock prevents concurrent runs. This is the primary consistency guarantee: it also catches mutations that bypass the save hooks (updateColumn() on paid/expiration/toped_until/deleted, recountPrices(), raw SQL), which the incremental sync cannot see.

Frequency: once a day, at night.

Runs in chunks (about 20 s per call, progress kept in Redis). Until finished it answers status: skipped (reason: reindex_continues); the scheduler calls the job again on the next tick, while a direct hosting cron must call it repeatedly until it returns status: ok. The sync of the same source finishes an unfinished reindex as a safety net.

URL: /cron/bazaar/elastic/reindex?token=xxx

Attributes
#[CronJob]
$frequency: 'daily:02:00'
$outbound: false

actionSync()

Incremental sync: indexes adverts inserted since the watermark and cleans up orphans.

public actionSync() : void

The watermark lives in Redis (es:<indexPrefix>:watermark:bazaar_advert, 5 min overlap); orphans are documents without a DB row. An advert has no update timestamp, so the sync catches only new adverts and orphaned documents: edits are propagated by the completedSave hook, payments and expiration by targeted triggers, everything else by the nightly full reindex.

Frequency: every 10 minutes.

URL: /cron/bazaar/elastic/sync?token=xxx

Attributes
#[CronJob]
$frequency: 'every:10'
$outbound: false

releaseLock()

Releases the lock at most once; called by both `shutdown()` and the PHP shutdown function.

public releaseLock() : void

sendResponse()

public sendResponse(Response $response) : void
Parameters
$response : Response

setConfigParameters()

public setConfigParameters([bool $debugMode = false ][, bool $productionMode = false ]) : void
Parameters
$debugMode : bool = false
$productionMode : bool = false

shutdown()

public shutdown(Response $response) : void
Parameters
$response : Response

isSameSiteRequest()

Tells whether the request comes from the own site (Nette SameSite cookie); protects GET actions with a side effect.

protected isSameSiteRequest() : bool

getHttpRequest() returns the IRequest interface, while isSameSite() exists only on Nette\Http\Request.

Return values
bool

sendError()

protected sendError(string $message[, int $code = 500 ]) : never
Parameters
$message : string
$code : int = 500
Return values
never

sendSkipped()

Reports that the job did only part of the work and the next call finishes the rest (e.g. a reindex done in chunks).

protected sendSkipped(string $reason[, array<string, mixed> $data = [] ]) : never

The scheduler does not record skipped as a run, so the job stays due and is called again on the next tick.

Parameters
$reason : string
$data : array<string, mixed> = []
Return values
never

sendSuccess()

protected sendSuccess([array<string|int, mixed> $data = [] ]) : never
Parameters
$data : array<string|int, mixed> = []
Return values
never

tryCall()

Runs the `action<Action>()` method as one operation (transitional).

protected tryCall(string $method, array<string, mixed> $params) : bool

Cron has no internalAction*(): its actions call the Api managers directly, so without this in scoped mode nothing would ever empty the unit of work of a cron run. The boundary is the action only; startup() (settings, token, lock) stays outside like the startup of an Api request. A cron action ends with AbortException from send*(), which is a regular end: the outcome comes from the response (ok succeeded, skipped unknown — possibly a part of a batch was done, error failed). Render and signal methods are not wrapped (cron presenters have none).

Parameters
$method : string
$params : array<string, mixed>
Return values
bool
On this page

Search results