Some more worker fine tuning

This commit is contained in:
Michael 2017-06-17 13:58:18 +00:00
parent d2f2e1c371
commit cf179188b6
7 changed files with 47 additions and 19 deletions

View file

@ -42,7 +42,7 @@ define ( 'FRIENDICA_PLATFORM', 'Friendica');
define ( 'FRIENDICA_CODENAME', 'Asparagus'); define ( 'FRIENDICA_CODENAME', 'Asparagus');
define ( 'FRIENDICA_VERSION', '3.5.3-dev' ); define ( 'FRIENDICA_VERSION', '3.5.3-dev' );
define ( 'DFRN_PROTOCOL_VERSION', '2.23' ); define ( 'DFRN_PROTOCOL_VERSION', '2.23' );
define ( 'DB_UPDATE_VERSION', 1229 ); define ( 'DB_UPDATE_VERSION', 1230 );
/** /**
* @brief Constant with a HTML line break. * @brief Constant with a HTML line break.

View file

@ -37,7 +37,6 @@ Example: To set the directory value please add this line to your .htconfig.php:
* **dbclean** (Boolean) - Enable the automatic database cleanup process * **dbclean** (Boolean) - Enable the automatic database cleanup process
* **dbclean-expire-days** (Integer) - Days after which remote items will be deleted. Own items, and marked or filed items are kept. * **dbclean-expire-days** (Integer) - Days after which remote items will be deleted. Own items, and marked or filed items are kept.
* **default_service_class** - * **default_service_class** -
* **delivery_batch_count** - Number of deliveries per process. Default value is 1. (Disabled when using the worker)
* **diaspora_test** (Boolean) - For development only. Disables the message transfer. * **diaspora_test** (Boolean) - For development only. Disables the message transfer.
* **directory** - The path to global directory. If not set then "http://dir.friendica.social" is used. * **directory** - The path to global directory. If not set then "http://dir.friendica.social" is used.
* **disable_email_validation** (Boolean) - Disables the check if a mail address is in a valid format and can be resolved via DNS. * **disable_email_validation** (Boolean) - Disables the check if a mail address is in a valid format and can be resolved via DNS.
@ -69,6 +68,7 @@ Example: To set the directory value please add this line to your .htconfig.php:
* **paranoia** (Boolean) - Log out users if their IP address changed. * **paranoia** (Boolean) - Log out users if their IP address changed.
* **permit_crawling** (Boolean) - Restricts the search for not logged in users to one search per minute. * **permit_crawling** (Boolean) - Restricts the search for not logged in users to one search per minute.
* **worker_debug** (Boolean) - If enabled, it prints out the number of running processes split by priority. * **worker_debug** (Boolean) - If enabled, it prints out the number of running processes split by priority.
* **worker_fetch_limit** - Number of worker tasks that are fetched in a single query. Default is 5.
* **profiler** (Boolean) - Enable internal timings to help optimize code. Needed for "rendertime" addon. Default is false. * **profiler** (Boolean) - Enable internal timings to help optimize code. Needed for "rendertime" addon. Default is false.
* **free_crawls** - Number of "free" searches when "permit_crawling" is activated (Default value is 10) * **free_crawls** - Number of "free" searches when "permit_crawling" is activated (Default value is 10)
* **crawl_permit_period** - Period in seconds between allowed searches when the number of free searches is reached and "permit_crawling" is activated (Default value is 60) * **crawl_permit_period** - Period in seconds between allowed searches when the number of free searches is reached and "permit_crawling" is activated (Default value is 60)

View file

@ -82,6 +82,9 @@ function cron_run(&$argv, &$argc){
proc_run(PRIORITY_MEDIUM, 'include/dbclean.php'); proc_run(PRIORITY_MEDIUM, 'include/dbclean.php');
proc_run(PRIORITY_LOW, "include/cronjobs.php", "update_photo_albums"); proc_run(PRIORITY_LOW, "include/cronjobs.php", "update_photo_albums");
// Delete all done workerqueue entries
dba::delete('workerqueue', array('done' => true));
} }
// Poll contacts // Poll contacts

View file

@ -1739,6 +1739,7 @@ function db_definition() {
"created" => array("type" => "datetime", "not null" => "1", "default" => NULL_DATE), "created" => array("type" => "datetime", "not null" => "1", "default" => NULL_DATE),
"pid" => array("type" => "int(11)", "not null" => "1", "default" => "0"), "pid" => array("type" => "int(11)", "not null" => "1", "default" => "0"),
"executed" => array("type" => "datetime", "not null" => "1", "default" => NULL_DATE), "executed" => array("type" => "datetime", "not null" => "1", "default" => NULL_DATE),
"done" => array("type" => "tinyint(1)", "not null" => "1", "default" => "0"),
), ),
"indexes" => array( "indexes" => array(
"PRIMARY" => array("id"), "PRIMARY" => array("id"),

View file

@ -18,7 +18,9 @@ if (!file_exists("boot.php") && (sizeof($_SERVER["argv"]) != 0)) {
require_once("boot.php"); require_once("boot.php");
function poller_run($argv, $argc){ function poller_run($argv, $argc){
global $a, $db; global $a, $db, $poller_up_start;
$poller_up_start = microtime(true);
$a = new App(dirname(__DIR__)); $a = new App(dirname(__DIR__));
@ -129,7 +131,7 @@ function poller_run($argv, $argc){
* @return integer Number of non executed entries in the worker queue * @return integer Number of non executed entries in the worker queue
*/ */
function poller_total_entries() { function poller_total_entries() {
$s = q("SELECT COUNT(*) AS `total` FROM `workerqueue` WHERE `executed` <= '%s'", dbesc(NULL_DATE)); $s = q("SELECT COUNT(*) AS `total` FROM `workerqueue` WHERE `executed` <= '%s' AND NOT `done`", dbesc(NULL_DATE));
if (dbm::is_result($s)) { if (dbm::is_result($s)) {
return $s[0]["total"]; return $s[0]["total"];
} else { } else {
@ -143,7 +145,7 @@ function poller_total_entries() {
* @return integer Number of active poller processes * @return integer Number of active poller processes
*/ */
function poller_highest_priority() { function poller_highest_priority() {
$s = q("SELECT `priority` FROM `workerqueue` WHERE `executed` <= '%s' ORDER BY `priority` LIMIT 1", dbesc(NULL_DATE)); $s = q("SELECT `priority` FROM `workerqueue` WHERE `executed` <= '%s' AND NOT `done` ORDER BY `priority` LIMIT 1", dbesc(NULL_DATE));
if (dbm::is_result($s)) { if (dbm::is_result($s)) {
return $s[0]["priority"]; return $s[0]["priority"];
} else { } else {
@ -159,7 +161,7 @@ function poller_highest_priority() {
* @return integer Is there a process running with that priority? * @return integer Is there a process running with that priority?
*/ */
function poller_process_with_priority_active($priority) { function poller_process_with_priority_active($priority) {
$s = q("SELECT `id` FROM `workerqueue` WHERE `priority` <= %d AND `executed` > '%s' LIMIT 1", $s = q("SELECT `id` FROM `workerqueue` WHERE `priority` <= %d AND `executed` > '%s' AND NOT `done` LIMIT 1",
intval($priority), dbesc(NULL_DATE)); intval($priority), dbesc(NULL_DATE));
return dbm::is_result($s); return dbm::is_result($s);
} }
@ -172,6 +174,7 @@ function poller_process_with_priority_active($priority) {
* @return boolean "true" if further processing should be stopped * @return boolean "true" if further processing should be stopped
*/ */
function poller_execute($queue) { function poller_execute($queue) {
global $poller_db_duration;
$a = get_app(); $a = get_app();
@ -212,9 +215,13 @@ function poller_execute($queue) {
if (function_exists($funcname)) { if (function_exists($funcname)) {
poller_exec_function($queue, $funcname, $argv); poller_exec_function($queue, $funcname, $argv);
dba::delete('workerqueue', array('id' => $queue["id"]));
$stamp = (float)microtime(true);
dba::update('workerqueue', array('done' => true), array('id' => $queue["id"]));
$poller_db_duration = (microtime(true) - $stamp);
} else { } else {
logger("Function ".$funcname." does not exist"); logger("Function ".$funcname." does not exist");
dba::delete('workerqueue', array('id' => $queue["id"]));
} }
return true; return true;
@ -228,6 +235,7 @@ function poller_execute($queue) {
* @param array $argv Array of values to be passed to the function * @param array $argv Array of values to be passed to the function
*/ */
function poller_exec_function($queue, $funcname, $argv) { function poller_exec_function($queue, $funcname, $argv) {
global $poller_up_start, $poller_db_duration;
$a = get_app(); $a = get_app();
@ -260,6 +268,8 @@ function poller_exec_function($queue, $funcname, $argv) {
$a->process_id = uniqid("wrk", true); $a->process_id = uniqid("wrk", true);
$a->queue = $queue; $a->queue = $queue;
$up_duration = number_format(microtime(true) - $poller_up_start, 3);
$funcname($argv, $argc); $funcname($argv, $argc);
$a->process_id = $old_process_id; $a->process_id = $old_process_id;
@ -267,6 +277,15 @@ function poller_exec_function($queue, $funcname, $argv) {
$duration = number_format(microtime(true) - $stamp, 3); $duration = number_format(microtime(true) - $stamp, 3);
$poller_up_start = microtime(true);
/** With these values we can analyze how effective the worker is.
* The database and rest time should be low since this is the unproductive time.
* The execution time is the productive time.
* By changing parameters like the maximum number of workers we can check the effectivness.
*/
logger('DB: '.number_format($poller_db_duration, 2).' - Rest: '.number_format($up_duration - $poller_db_duration, 2).' - Execution: '.number_format($duration, 2), LOGGER_DEBUG);
if ($duration > 3600) { if ($duration > 3600) {
logger("Prio ".$queue["priority"].": ".$queue["parameter"]." - longer than 1 hour (".round($duration/60, 3).")", LOGGER_DEBUG); logger("Prio ".$queue["priority"].": ".$queue["parameter"]." - longer than 1 hour (".round($duration/60, 3).")", LOGGER_DEBUG);
} elseif ($duration > 600) { } elseif ($duration > 600) {
@ -419,7 +438,7 @@ function poller_max_connections_reached() {
* *
*/ */
function poller_kill_stale_workers() { function poller_kill_stale_workers() {
$r = q("SELECT `pid`, `executed`, `priority`, `parameter` FROM `workerqueue` WHERE `executed` > '%s'", dbesc(NULL_DATE)); $r = q("SELECT `pid`, `executed`, `priority`, `parameter` FROM `workerqueue` WHERE `executed` > '%s' AND NOT `done`", dbesc(NULL_DATE));
if (!dbm::is_result($r)) { if (!dbm::is_result($r)) {
// No processing here needed // No processing here needed
@ -493,16 +512,16 @@ function poller_too_much_workers() {
$listitem = array(); $listitem = array();
// Adding all processes with no workerqueue entry // Adding all processes with no workerqueue entry
$processes = dba::p("SELECT COUNT(*) AS `running` FROM `process` WHERE NOT EXISTS (SELECT id FROM `workerqueue` WHERE `workerqueue`.`pid` = `process`.`pid`)"); $processes = dba::p("SELECT COUNT(*) AS `running` FROM `process` WHERE NOT EXISTS (SELECT id FROM `workerqueue` WHERE `workerqueue`.`pid` = `process`.`pid` AND NOT `done`)");
if ($process = dba::fetch($processes)) { if ($process = dba::fetch($processes)) {
$listitem[0] = "0:".$process["running"]; $listitem[0] = "0:".$process["running"];
} }
dba::close($processes); dba::close($processes);
// Now adding all processes with workerqueue entries // Now adding all processes with workerqueue entries
$entries = dba::p("SELECT COUNT(*) AS `entries`, `priority` FROM `workerqueue` GROUP BY `priority`"); $entries = dba::p("SELECT COUNT(*) AS `entries`, `priority` FROM `workerqueue` AND NOT `done` GROUP BY `priority`");
while ($entry = dba::fetch($entries)) { while ($entry = dba::fetch($entries)) {
$processes = dba::p("SELECT COUNT(*) AS `running` FROM `process` INNER JOIN `workerqueue` ON `workerqueue`.`pid` = `process`.`pid` WHERE `priority` = ?", $entry["priority"]); $processes = dba::p("SELECT COUNT(*) AS `running` FROM `process` INNER JOIN `workerqueue` ON `workerqueue`.`pid` = `process`.`pid` AND NOT `done` WHERE `priority` = ?", $entry["priority"]);
if ($process = dba::fetch($processes)) { if ($process = dba::fetch($processes)) {
$listitem[$entry["priority"]] = $entry["priority"].":".$process["running"]."/".$entry["entries"]; $listitem[$entry["priority"]] = $entry["priority"].":".$process["running"]."/".$entry["entries"];
} }
@ -564,7 +583,7 @@ function poller_passing_slow(&$highest_priority) {
$r = q("SELECT `priority` $r = q("SELECT `priority`
FROM `process` FROM `process`
INNER JOIN `workerqueue` ON `workerqueue`.`pid` = `process`.`pid`"); INNER JOIN `workerqueue` ON `workerqueue`.`pid` = `process`.`pid` AND NOT `done`");
// No active processes at all? Fine // No active processes at all? Fine
if (!dbm::is_result($r)) { if (!dbm::is_result($r)) {
@ -609,12 +628,13 @@ function find_worker_processes() {
// Check if we should pass some low priority process // Check if we should pass some low priority process
$highest_priority = 0; $highest_priority = 0;
$found = false; $found = false;
$limit = Config::get('system', 'worker_fetch_limit', 5);
if (poller_passing_slow($highest_priority)) { if (poller_passing_slow($highest_priority)) {
// Are there waiting processes with a higher priority than the currently highest? // Are there waiting processes with a higher priority than the currently highest?
$result = dba::e("UPDATE `workerqueue` SET `executed` = ?, `pid` = ? $result = dba::e("UPDATE `workerqueue` SET `executed` = ?, `pid` = ?
WHERE `executed` <= ? AND `priority` < ? WHERE `executed` <= ? AND `priority` < ? AND NOT `done`
ORDER BY `priority`, `created` LIMIT 5", ORDER BY `priority`, `created` LIMIT ".intval($limit),
datetime_convert(), getmypid(), NULL_DATE, $highest_priority); datetime_convert(), getmypid(), NULL_DATE, $highest_priority);
if ($result) { if ($result) {
$found = (dba::affected_rows() > 0); $found = (dba::affected_rows() > 0);
@ -623,7 +643,7 @@ function find_worker_processes() {
if (!$found) { if (!$found) {
// Give slower processes some processing time // Give slower processes some processing time
$result = dba::e("UPDATE `workerqueue` SET `executed` = ?, `pid` = ? $result = dba::e("UPDATE `workerqueue` SET `executed` = ?, `pid` = ?
WHERE `executed` <= ? AND `priority` > ? WHERE `executed` <= ? AND `priority` > ? AND NOT `done`
ORDER BY `priority`, `created` LIMIT 1", ORDER BY `priority`, `created` LIMIT 1",
datetime_convert(), getmypid(), NULL_DATE, $highest_priority); datetime_convert(), getmypid(), NULL_DATE, $highest_priority);
if ($result) { if ($result) {
@ -634,7 +654,7 @@ function find_worker_processes() {
// If there is no result (or we shouldn't pass lower processes) we check without priority limit // If there is no result (or we shouldn't pass lower processes) we check without priority limit
if (!$found) { if (!$found) {
$result = dba::e("UPDATE `workerqueue` SET `executed` = ?, `pid` = ? WHERE `executed` <= ? ORDER BY `priority`, `created` LIMIT 5", $result = dba::e("UPDATE `workerqueue` SET `executed` = ?, `pid` = ? WHERE `executed` <= ? AND NOT `done` ORDER BY `priority`, `created` LIMIT ".intval($limit),
datetime_convert(), getmypid(), NULL_DATE); datetime_convert(), getmypid(), NULL_DATE);
if ($result) { if ($result) {
$found = (dba::affected_rows() > 0); $found = (dba::affected_rows() > 0);
@ -649,12 +669,13 @@ function find_worker_processes() {
* @return string SQL statement * @return string SQL statement
*/ */
function poller_worker_process() { function poller_worker_process() {
global $poller_db_duration;
$stamp = (float)microtime(true); $stamp = (float)microtime(true);
$found = find_worker_processes(); $found = find_worker_processes();
logger('Duration: '.number_format(microtime(true) - $stamp, 3), LOGGER_DEBUG); $poller_db_duration += (microtime(true) - $stamp);
if ($found) { if ($found) {
$r = q("SELECT * FROM `workerqueue` WHERE `pid` = %d", intval(getmypid())); $r = q("SELECT * FROM `workerqueue` WHERE `pid` = %d", intval(getmypid()));

View file

@ -17,9 +17,12 @@ function display_init(App $a) {
if (substr($a->argv[2], -5) == '.atom') { if (substr($a->argv[2], -5) == '.atom') {
$item_id = substr($a->argv[2], 0, -5); $item_id = substr($a->argv[2], 0, -5);
$xml = dfrn::itemFeed($item_id); $xml = dfrn::itemFeed($item_id);
if ($xml == '') {
http_status_exit(500);
}
header("Content-type: application/atom+xml"); header("Content-type: application/atom+xml");
echo $xml; echo $xml;
http_status_exit(($xml) ? 200 : 500); killme();
} }
} }

View file

@ -1,6 +1,6 @@
<?php <?php
define('UPDATE_VERSION' , 1229); define('UPDATE_VERSION' , 1230);
/** /**
* *