#!/usr/bin/env perl
use 5.008;
use strict;
use warnings;
use FindBin ();
use File::Basename ();
use File::Spec;
use Getopt::Long;
use POSIX;
use Sys::Hostname;
# [LH] 2013-08-04: Added Fcntl for better pidfile locking. [RT81914]
use Fcntl qw(:DEFAULT :flock);
use Error qw(:try);
use Helios;
use Helios::Error;
use Helios::LogEntry::Levels qw(:all);
# [LH] [2013-10-04]: Switched to using Helios::TS for virtual jobtypes.
use Helios::TS;
use Helios::Config;
our $VERSION = '2.71_4051';
# FILE CHANGE HISTORY
# [2012-01-08]: Added a check to try to prevent loading code outside of @INC.
# [2012-01-09]: Modified check to prevent loading code outside of @INC.
# [2012-03-25]: Added code to prevent a worker process from changing into a
# daemon due to odd (and uncommon) database connection instability.
# [2012-03-27]: Changed $DEFAULTS{ZERO_SLEEP_INTERVAL} to 10. Added code to
# add "registration_interval" as a configuration parameter. Added
# $DEFAULTS{REGISTRATION_INTERVAL} and set to 60. Changed
# $REGISTRATION_INTERVAL to 60.
# [2012-04-01]: Changed service module loaded notification to report version
# only if the module has a $VERSION set. Changed service initialization to use
# the prep() method instead of getConfigFromIni() and getConfigFromDb() to
# ensure proper initialization of loggers and Data::ObjectDriver database
# connection.
# [2012-05-20]: Removed old commented out calls to getConfigFrom*(),
# clean_shutdown(), and logMsg() (logging a HALT message). Removed 2 empty
# comments.
# [LH] [2012-07-11]: Switched from using TheSchwartz to using
# Helios::TheSchwartz. Added WORKER_BLITZ_FACTOR feature to launch workers
# faster when there are less than MAX_WORKERS jobs available. Stopped closing
# STDOUT, STDIN, and STDERR when helios.pl daemonizes because it was causing
# problems with some services.
# [LH] [2012-07-15]: Added to use Helios::Config, and changed configuration
# update code to use Helios::Config instead of getConfigFromDb(). Added
# debugging messages for ZERO_SLEEP_INTERVAL and REGISTRATION_INTERVAL.
# [LH] [2012-07-26]: Changed "worker_blitz_factor" config param to
# "WORKER_BLITZ_FACTOR". Added new code to better handle database connections
# after a fork() so worker processes do not share or disconnect the daemon's
# connections.
# [LH] [2012-07-29]: Added DOUBLE_CLUTCH_INTERVAL config parameter to control
# WORKER_MAX_TTL functionality better than ZERO_LAUNCH_INTERVAL. Updated
# copyright info.
# [LH] [2012-09-05]: Changed DOUBLE_CLUTCH_INTERVAL to
# WORKER_MAX_TTL_WAIT_INTERVAL.
# [LH] [2012-09-28]: Removed old commented out code.
# [LH] [2012-10-12]: Replaced code attempting to prevent loading of modules
# outside of @INC with new require_module() function.
# [LH] [2012-12-11]: Added code to apply WORKER_MAX_TTL after registration in
# daemon main loop. [RT81709]
# [LH] [2013-08-04]: Implemented new PID file locking scheme to fix [RT81914].
# Replaced write_pid_file() and mostly rewrote running_process_check() to
# prevent a PID file race condition.
# [LH] [2013-08-19]: Added comments for clarification of fix for [RT81914].
# [LH] [2013-09-21]: Added code to enable job prioritization features in
# Helios::TheSchwartz. Added code to implement WORKER_LAUNCH_PATTERN feature.
# [LH] [2013-10-04]: Added code to implement "virtual jobtypes." Switched to
# using Helios::TS instead of Helios::TheSchwartz. Changed command line
# option parsing to accomodate options other than "--clear-halt". Removed
# old commented out code for class name sanity checking and class loading.
# Removed old code to calculate workers to launch, also removed code for
# WORKER_BLITZ_FACTOR feature (WORKER_LAUNCH_PATTERN replaces it).
=head1 NAME
helios.pl - Launch a daemon to service jobs in the Helios job processing system
=head1 SYNOPSIS
export HELIOS_INI=/path/to/helios.ini
[export HELIOS_DEBUG=1]
helios.pl <jobclass> [--clear-halt]
# just prints version info
helios.pl --version
=head1 DESCRIPTION
The helios.pl program, given a subclass of Helios::Service, will launch a daemon to service Helios
jobs of that subclass. The number of worker processes to run concurrently and various other
parameters are set via a helios.ini file and the Helios database (the connection information
of which is also defined in helios.ini).
Under normal operation, helios.pl will attempt to load the service class specified on the command
line and use it to read the contents of the helios.ini file. If successful, it will attempt to
connect to the Helios collective database and read the relevant configuration parameters from
there. If that is successful, helios.pl will daemonize and start servicing jobs of the specified
class.
In debug mode (set HELIOS_DEBUG=1), helios.pl will not disconnect from the terminal, and will
output extra debugging information to the screen and the Helios log. It will also enable debug mode
on the service class, which may also support the output of extra debugging information.
If the --clear-halt option is specified, helios.pl will attempt to remove a HALT
parameter specified in the Helios configuration for the specified service on the
current host. This is helpful if you shutdown your Helios service daemon using
the Helios::Panoptes Collective Admin view. Note that it will NOT remove a HALT
specified globally (where host = '*').
=head1 HELIOS.INI
The initial parameters for helios.pl are defined in an INI-style configuration file typically
named "helios.ini". The file's location is normally specified by setting the HELIOS_INI
environment variable before helios.pl is started. If HELIOS_INI is not set, helios.pl will default
to the helios.ini in the current directory, if present.
A helios.ini file normally contains a [global] section with the parameters
necessary to connect to the Helios collective database, and any parameters local
to the host on which the Helios service is currently running. In addition,
each service class may have its own section in helios.ini containing
parameters specfic to that service on that host.
Example helios.ini:
[global]
dsn=dbi:mysql:host=10.1.0.21;db=helios_db
user=helios
password=password
[LongRunningJobService]
master_launch_interval=60
zero_launch_interval=90
[SearchIndex::LoadTestService]
OVERDRIVE=1
MAX_WORKERS=3
=head2 Helios.ini Parameters for helios.pl
=head3 Configuration options to place in the [global] section:
=over 4
=item dsn
Datasource name for the Helios parameter database. This will also contain the data structures for
the underlying TheSchwartz queuing system.
=item user
Database user for the datasource name described above.
=item password
Database password for the datasource name described above.
=item pid_path
The location where helios.pl should write the PID file for this service
instance. The default is "/var/run/helios". The name of the PID file will be
a variation on the service class's name.
=back
=head3 Configuration options to place in individual service sections:
=over 4
=item master_launch_interval
Set the master_launch_interval to determine how long helios.pl should sleep after launching
workers before accessing the database to update its configuration parameters and check for waiting
jobs. The default is 1 second which works well for normally short-lived jobs (2 secs or less),
but that may be overkill for longer-lived jobs (jobs that run longer than 2 sec). Setting the
master_launch_interval option may help prevent needless database traffic for longer running jobs.
=item zero_launch_interval
Set the zero_launch_interval to determine how long helios.pl should sleep after reaching its
MAX_WORKERS limit. The default is 10 sec. If jobs are running long enough that
helios.pl is frequently hitting is MAX_WORKERS limit (there are waiting jobs but
helios.pl can't launch new workers because previously launched jobs are still running), increasing
the zero_launch_interval will reduce needless database traffic.
=back
=head1 HELIOS CTRL PANEL (helios_params_tb)
In addition to helios.ini, certain helios.pl configuration options can be set via the Ctrl Panel
in the Helios::Panoptes web interface. These configuration options are read by helios.pl from the
HELIOS_PARAMS_TB table in the Helios database. Though the Ctrl Panel is designed mostly to
provide a standardized interface for service classes to store and retrieve configuration
parameters, certain helios.pl control parameters can be set here as well.
The helios.pl daemon will refresh its configuration parameters from the database after waiting
master_launch_interval seconds after launching worker processes. If MAX_WORKERS
workers are still running, the helios.pl daemon will wait zero_launch_interval seconds to refresh
its configuration parameters and try to launch workers again. If you have long running jobs (jobs
that last more than a few seconds), resetting these parameters higher may help smooth processing
and reduce needless database traffic. See the HELIOS OPERATION section below for more
information.
=head2 helios.pl Control Parameters in Ctrl Panel
=over 4
=item MAX_WORKERS
The Helios system is designed to run one daemon process for each service class per host. However,
each service daemon can run more than 1 worker process simultaneously. The upper limit for worker
processes is set using the MAX_WORKERS option. The helios.pl default is 1 worker per
server. You can provide a different default for a particular class by specifying a value in
helios.ini. If active management of available workers is not necessary for a
particular service class, this option can be set in helios.ini.
=item OVERDRIVE
Speed up job processing by allowing workers to service jobs until no more are in the queue.
Normally, a worker process will service a single job and then exit (1:1 worker/job ratio). With
OVERDRIVE mode enabled, a worker process will keep servicing jobs until it can find no more of
that type in the job queue (1:many worker/job ratio). This may help to speed up processing for
short-lived jobs, as once a worker process has started, it will stay in memory, continuing to
process jobs until there are no more to process. This can also enable on-the-fly caching of
resources, as once a worker has loaded a resource or established a database connection for a job
it can be held for use by subsequent jobs, potentially reducing database load or disk I/O. Longer
running jobs may not benefit as much, however.
Like MAX_WORKERS, if active management of the job processing mode is not necessary, this
parameter can be set in helios.ini (though that is discouraged).
=item HOLD
If set to 1, a HOLD parameter will cause the helios.pl service daemon to stop launching new workers
to process jobs. Currently processing jobs will be allowed to complete, and those worker
processes will end normally.
Setting HOLD to 0 will allow normal job processing to resume.
=item HALT
If a HALT parameter is defined, the helios.pl service daemon will start process cleanup and will
exit. As with HOLD = 1, jobs currently processing will complete, and the worker processes
will exit normally.
Once a helios.pl daemon has HALTED, there is no way to restart it from the Helios::Pantoptes
interface. It must be restarted from the command line with a 'helios.pl <serviceclass>' command.
As of Helios 2.22, service daemons can also be shut down in the normal *nix manner of using
the 'kill' command to send the daemon a SIGTERM signal.
=item WORKER_MAX_TTL
The time in seconds to allow a worker to run. If you have a problem with
workers that get stuck while performing a job (perhaps waiting for a database
connection, or some other problem), you can set WORKER_MAX_TTL for that service.
The helios.pl program will check to see if workers for that service are
running beyond their intended time-to-live, and kill them if they run too long.
=back
=cut
# for SIGHUP support
our @ORIGINAL_ARGV = @ARGV;
my $basename = File::Basename::basename($0);
our $HELIOS_PL_PATH = File::Spec->catfile($FindBin::Bin, $basename);
# globals settings
our $DEBUG_MODE = $ENV{HELIOS_DEBUG};
our $HELIOS_INI = $ENV{HELIOS_INI};
# BEGIN CODE Copyright (C) 2013 by Logical Helion, LLC.
# [LH] [2013-10-04]: Virtual jobtypes. We have to parse the cmd line args
# in a completely different way, but still maintain backward compatibility.
# if the first cmd line switch starts with '-', then all the options are
# "proper." If not, then the first option is the service class, and the
# remaining options will be proper.
our $CLASS;
our $OPT_CLEAR_HALT = 0;
our $OPT_HELP = 0;
our $OPT_JOBTYPES = '';
our $OPT_VERSION = 0;
if ($ARGV[0] !~ /^\-/) {
$CLASS = shift @ARGV;
}
GetOptions(
"service=s" => \$CLASS,
"clear-halt" => \$OPT_CLEAR_HALT,
"help" => \$OPT_HELP,
"jobtypes=s" => \$OPT_JOBTYPES,
"version" => \$OPT_VERSION,
);
$OPT_JOBTYPES =~ s/ //g; #[]more sanity checking!
our @ALT_JOBTYPES = split(/,/, $OPT_JOBTYPES);
# END CODE Copyright (C) 2013 by Logical Helion, LLC.
# other globals
# max workers will default to 1 if not set elsewhere
our %DEFAULTS = (
MAX_WORKERS => 1,
HELIOS_INI => File::Spec->catfile(File::Spec->curdir,'helios.ini'),
PID_PATH => File::Spec->catfile(File::Spec->rootdir, 'var', 'run', 'helios'),
MASTER_LAUNCH_INTERVAL => 1,
ZERO_LAUNCH_INTERVAL => 10,
ZERO_SLEEP_INTERVAL => 10,
REGISTRATION_INTERVAL => 60,
WORKER_BLITZ_FACTOR => 1,
WORKER_MAX_TTL_WAIT_INTERVAL => 20,
PRIORITIZE_JOBS => 0,
WORKER_LAUNCH_PATTERN => 'cons',
);
our $CLEAN_SHUTDOWN = 1; # used to determine if we should remove the PID file or not (at least for now)
our $HOLD_LOG_INTERVAL = 3600; # to reduce needless log msg in log_tb while HOLDing
our $HOLD_LOG_LAST = 0;
our $MASTER_LAUNCH_INTERVAL;
our $ZERO_LAUNCH_INTERVAL;
our $START_TIME = time(); # used to measure uptime from the registry table
our $REGISTRATION_INTERVAL = 60; # used to periodically register daemon in database
our $REGISTRATION_LAST = 0;
our $PID_FILE; # globally accessible PID file location
# [LH] 2013-08-04: Added $PID_FILE_H as a filehandle for better pidfile locking. [RT81914]
our $PID_FILE_H; # globally accessible PID file handle
our $SAFE_MODE_DELAY = 45; # SAFE MODE support; number of secs to wait
our $SAFE_MODE_RETRIES = 5; # SAFE MODE support; number of times to retry
our $ZERO_SLEEP_INTERVAL; # to reduce needless checking of the database
our $ZERO_SLEEP_LOG_INTERVAL = 3600; # to reduce needless log msgs in log_tb
our $ZERO_SLEEP_LOG_LAST = 0;
our $WORKER_MAX_TTL_WAIT_INTERVAL = 20; # for WORKER_MAX_TTL functionality
our $WORKER_PROCESS = 0; # used to indicate process has become a worker process
# this is used in addition to getppid() to prevent workers
# from becoming daemons in case of database instability
our $WORKER_BLITZ_FACTOR = 1; # used to determine how many workers to launch
our $HELIOS_DB_CONN; # database handle to help with DBI+forking issues
# print help if asked
# [LH] [2013-10-04]: New @ARGV parsing for "--help".
if ( !defined($CLASS) || $OPT_HELP ) {
require Pod::Usage;
Pod::Usage::pod2usage(-verbose => 2, -exitstatus => 0);
}
# conditionally load module or die in the attempt
my $worker_class = $CLASS;
print "Helios ",$Helios::VERSION,"\n";
print "helios.pl Service Daemon version $VERSION\n";
# --version support
# BEGIN CODE Copyright (C) 2013 by Logical Helion, LLC.
# [LH] [2013-10-04]: New @ARGV parsing for "--version".
if ( $OPT_VERSION ) { exit(0); }
# END CODE Copyright (C) 2013 by Logical Helion, LLC.
print "Attempting to load $worker_class...\n";
unless ( $worker_class ) {
print "You must specify the module name of a Helios service.\n";
exit(1);
}
require_module($worker_class);
if ( defined($worker_class->VERSION) ) {
print $worker_class, ' ', $worker_class->VERSION," loaded.\n";
} else{
print $worker_class," loaded.\n";
}
if ($DEBUG_MODE) { print "Debug Mode enabled.\n"; }
# instantiate a worker to access the necessary settings
my $worker = new $worker_class;
$worker->setHostname(hostname);
if ( defined($HELIOS_INI) ) {
$worker->setIniFile( $HELIOS_INI );
} else {
$worker->setIniFile( $DEFAULTS{HELIOS_INI} );
}
$worker->setJobType($worker_class);
# BEGIN CODE Copyright (C) 2013 by Logical Helion, LLC.
# [LH] [2013-10-04]: Virtual jobtypes set & init.
if (@ALT_JOBTYPES) {
$worker->setAltJobTypes(@ALT_JOBTYPES);
$worker->lookupAltJobtypeids();
}
# END CODE Copyright (C) 2013 by Logical Helion, LLC.
$worker->debug($DEBUG_MODE);
eval {
$worker->prep();
};
if ($@) {
die("FAILED to get configuration: $@");
}
my $params = $worker->getConfig();
if ($DEBUG_MODE) {
print "--INITIAL PARAMS--\n";
foreach my $param (keys %$params) {
print $param, ":", $params->{$param},"\n";
}
}
# SETUP OPERATIONAL PARAMETERS
if ( defined($params->{master_launch_interval}) ) {
$MASTER_LAUNCH_INTERVAL = $params->{master_launch_interval};
} else {
$MASTER_LAUNCH_INTERVAL = $DEFAULTS{MASTER_LAUNCH_INTERVAL};
}
if ( defined($params->{zero_launch_interval}) ) {
$ZERO_LAUNCH_INTERVAL = $params->{zero_launch_interval};
} else {
$ZERO_LAUNCH_INTERVAL = $DEFAULTS{ZERO_LAUNCH_INTERVAL};
}
if ( defined($params->{zero_sleep_interval}) ) {
$ZERO_SLEEP_INTERVAL = $params->{zero_sleep_interval};
} else {
$ZERO_SLEEP_INTERVAL = $DEFAULTS{ZERO_SLEEP_INTERVAL};
}
if ( defined($params->{registration_interval}) ) {
$REGISTRATION_INTERVAL = $params->{registration_interval};
} else {
$REGISTRATION_INTERVAL = $DEFAULTS{REGISTRATION_INTERVAL};
}
if ( defined($params->{WORKER_BLITZ_FACTOR}) ) {
$WORKER_BLITZ_FACTOR = $params->{WORKER_BLITZ_FACTOR};
} else {
$WORKER_BLITZ_FACTOR = $DEFAULTS{WORKER_BLITZ_FACTOR};
}
if ( defined($params->{WORKER_MAX_TTL_WAIT_INTERVAL}) ) {
$WORKER_MAX_TTL_WAIT_INTERVAL = $params->{WORKER_MAX_TTL_WAIT_INTERVAL};
} else {
$WORKER_MAX_TTL_WAIT_INTERVAL = $DEFAULTS{WORKER_MAX_TTL_WAIT_INTERVAL};
}
# make a globally accessible database handle
# to make it easier to clean up CachedKids after a fork()
$HELIOS_DB_CONN = $worker->dbConnect();
if ($DEBUG_MODE) {
print "MASTER LAUNCH INTERVAL: $MASTER_LAUNCH_INTERVAL\n";
print "ZERO LAUNCH INTERVAL: $ZERO_LAUNCH_INTERVAL\n";
print "ZERO SLEEP_INTERVAL: $ZERO_SLEEP_INTERVAL\n";
print "REGISTRATION_INTERVAL: $REGISTRATION_INTERVAL\n";
print "WORKER BLITZ FACTOR: $WORKER_BLITZ_FACTOR\n";
}
my %workers;
my $pid;
my $waiting_jobs = 0;
my $running_workers = 0;
my $max_workers = $DEFAULTS{MAX_WORKERS};
our $DATABASES_INFO = [
{ dsn => $params->{dsn},
user => $params->{user},
pass => $params->{password}
}
];
my $times_thru_loop = 0;
my $times_waiting = 0;
my $times_sleeping = 0;
# attempt to clear a HALT parameter, if specified
# then we'll check to see if HALT is still there
# if it is, we'll have to stop here rather than
# daemonize and launch the main loop
# [LH] [2013-10-04]: New @ARGV parsing for "--clear-halt".
if ( $OPT_CLEAR_HALT ) {
clear_halt();
$worker->prep();
$params = $worker->getConfig();
}
if ( defined($params->{HALT}) ) {
print STDERR "HALT is set for this service class.\n";
print STDERR "Please clear it and try again.\n";
print STDERR $worker_class," HALTED.\n";
exit(1);
}
# final check before we launch:
# check to make sure a daemon for this service isn't already running
if ( running_process_check($params->{pid_path}) ) {
print STDERR $worker->errstr(),"\n";
exit(1);
}
# Daemonize unless we're in debug mode
unless ($DEBUG_MODE) {
daemonize();
} else {
# print debug message
print "Writing pid file...\n";
write_pid_file($params->{pid_path}) or die($worker->errstr);
print "Executing main loop...\n";
}
# set up signal handler to reap dead children
$SIG{CHLD} = \&reaper;
$SIG{TERM} = \&terminator;
=head1 HELIOS OPERATION
After initial setup, the helios.pl daemon will enter a main operation loop where configuration
parameters are refreshed, the job queue is checked, and worker processes are launched and
cleaned up after. A HOLD = 1 parameter will temporarily cause the loop to pause processing, while
a HALT parameter will cause the helios.pl daemon to exit the loop, clean up, and exit.
There are several steps in the helios.pl main operation loop:
=over 4
=item 1.
Refresh configuration parameters from database.
=item 2.
Check job queue to see if there are jobs available for processing (if not, sleep for
zero_sleep_interval seconds and start again).
=item 3.
If there are jobs available, check to see how many worker processes are currently running.
If MAX_WORKERS workers are already running, sleep for zero_launch_interval seconds and
start again. The zero_launch_interval setting should be long enough to allow at least some of
the running jobs to complete (the default is 10 secs).
=item 4.
Subtract the number of currently running workers from MAX_WORKERS and launch that
many workers to handle the available jobs. If MAX_WORKERS is 5 and only 2
workers are running, that means the helios.pl daemon will launch 3 workers. If
there are less than MAX_WORKERS jobs available for processing, only 1 worker
will be launched to reduce job contention between workers in the collective.
=item 5.
Sleep master_launch_interval seconds and start the operation loop again.
=back
=cut
MAIN_LOOP:{
try {
# while not halted
while (!defined($params->{HALT}) ) {
$times_thru_loop++;
# recheck db parameters
$params = undef;
$params = Helios::Config->parseConfig();
# DAEMON REGISTRATION
# every $REGISTRATION_INTERVAL seconds, (re)register this daemon in the database
if ( ($REGISTRATION_LAST + $REGISTRATION_INTERVAL) < time() ) {
register();
$REGISTRATION_LAST = time();
# [LH] 2012-12-11: Copied WORKER_MAX_TTL/double_clutch() call from HOLD code below
# to enable WORKER_MAX_TTL in Normal Mode as well as Hold Mode. [RT81709]
if ( defined($params->{WORKER_MAX_TTL}) && $params->{WORKER_MAX_TTL} > 0
&& scalar(keys %workers) ) {
double_clutch();
}
}
# HOLDING JOB PROCESSING
# hold launching jobs temporarily
if ( defined($params->{HOLD}) && ($params->{HOLD} == 1) ) {
if ( $HOLD_LOG_LAST + $HOLD_LOG_INTERVAL < time() ) {
$worker->logMsg(LOG_NOTICE, "$0 $worker_class HOLDING");
$HOLD_LOG_LAST = time();
}
if ($DEBUG_MODE) { print "$0 $worker_class HOLDING\n"; }
sleep 60;
# after the first cycle through HOLD, the workers had BETTER be dead
if ( defined($params->{WORKER_MAX_TTL}) && $params->{WORKER_MAX_TTL} > 0
&& scalar(keys %workers) ) {
double_clutch();
}
next;
}
# once we're not holding anymore, we'll want to log the next time we enter HOLD mode
$HOLD_LOG_LAST = 0;
# DETERMINING WORKERS TO LAUNCH
$waiting_jobs = $worker->jobsWaiting();
$running_workers = scalar(keys %workers);
$max_workers = defined($params->{MAX_WORKERS}) ? $params->{MAX_WORKERS} : $DEFAULTS{MAX_WORKERS};
if ($DEBUG_MODE) { $worker->logMsg(LOG_DEBUG, "MAX WORKERS: $max_workers"); }
if ($DEBUG_MODE) { $worker->logMsg(LOG_DEBUG, "RUNNING WORKERS: $running_workers"); }
# if no waiting jobs, sleep
if ($DEBUG_MODE) {
print $waiting_jobs, " waiting ",$worker_class, " jobs.\n";
$worker->logMsg(LOG_DEBUG, $waiting_jobs . " waiting " . $worker_class . " jobs.");
}
unless ( $waiting_jobs ) {
$times_sleeping++;
# only log the "0 workers running, 0 workers in queue. SLEEPING" message every $ZERO_SLEEP_LOG_INTERVAL seconds
# (necessary to prevent overwhelming database logging with messages we don't care about)
if ( ($running_workers == 0) && (($ZERO_SLEEP_LOG_LAST + $ZERO_SLEEP_LOG_INTERVAL ) < time()) ) {
$worker->logMsg(LOG_NOTICE, $running_workers." workers running, ".$waiting_jobs." in queue. SLEEPING");
$ZERO_SLEEP_LOG_LAST = time();
}
sleep $ZERO_SLEEP_INTERVAL;
next;
}
# once we're not zero sleeping ("0 workers running, 0 workers in queue")
# we'll want to log the next time we go into zero sleep
$ZERO_SLEEP_LOG_LAST = 0;
# LAUNCHING WORKERS
# if we've got to this part of the loop, we have jobs that need workers to launch
# (though we still may not do it if we've already reached our limit)
# [LH] [2013-09-21]: Added code to implement WORKER_LAUNCH_PATTERN feature.
my $workers_to_launch = workers_to_launch($waiting_jobs, $running_workers, $max_workers);
$worker->logMsg(LOG_NOTICE, "$waiting_jobs jobs waiting; $running_workers workers running; launching $workers_to_launch workers");
for (my $i = 0; $i < $workers_to_launch; $i++) {
FORK: {
if ($pid = fork) {
# I'm the parent!
$workers{$pid} = time();
if ($DEBUG_MODE) { $worker->logMsg(LOG_NOTICE, $worker->getJobType()." process $pid launched"); }
sleep 1;
} elsif (defined $pid) { # $pid is zero here if defined
# I'm the child!
$WORKER_PROCESS = 1;
# BEGIN CODE Copyright (C) 2012 by Logical Helion, LLC.
# BEFORE WE LAUNCH THE WORKER,
# clean up the database connections from the parent
# we'll set InactiveDestroy on the existing connections
# then we'll clear them, leaving the parent connections
# free of any influence of the children
# NOTE: with DBI >= 1.614, all we'd have to do is
# set AutoInactiveDestroy on all the connections
# but to support the DBI (1.52) bundled with RHEL & CentOS 5,
# we have to make do with what we have
my $ck = $HELIOS_DB_CONN->{Driver}->{CachedKids};
foreach (keys %$ck) {
$ck->{$_}->{InactiveDestroy} = 1;
}
%$ck = ();
# END CODE Copyright (C) 2012 by Logical Helion, LLC.
# NOW, launch the worker
launch_worker();
} elsif ($! == EAGAIN) {
# EAGAIN is the supposedly recoverable fork error
sleep 2;
redo FORK;
} else {
# weird fork error
die "Can't fork: $!\n";
}
}
}
# SLEEP INTERVAL AFTER LAUNCHING
# depending on the worker type, we may need to wait before we go back to the top of the loop
# (pull new parameters and waiting job count from the database)
# This can help reduce needless database traffic
# if there are running jobs but we didn't launch any
# (max workers has been reached) sleep for a short while
# to reduce needless "spinning" (and db access)
if ( $workers_to_launch <= 0 ) {
if ( $DEBUG_MODE ) { $worker->logMsg(LOG_DEBUG, "MAX WORKERS REACHED, SLEEPING"); }
sleep $ZERO_LAUNCH_INTERVAL;
if ( defined($params->{WORKER_MAX_TTL}) && $params->{WORKER_MAX_TTL} > 0 ) {
double_clutch();
}
next;
}
# if we did launch jobs, depending on the type of worker we may not want to re-check the database yet
# longer-running jobs (eg Helios::BatchWorker) don't need to have the db checked immediately after launch
# shorter-running jobs (eg IndexWorker) have likely already completed in the time it took to launch them
# and we want to launch as many as possible as quickly as possible (esp. if there's a Mass Index operation)
if ($MASTER_LAUNCH_INTERVAL) {
if ( $DEBUG_MODE ) { $worker->logMsg(LOG_DEBUG, "MASTER LAUNCH INTERVAL $MASTER_LAUNCH_INTERVAL, SLEEPING"); }
sleep $MASTER_LAUNCH_INTERVAL;
}
}
} otherwise {
my $e = shift;
# if we're a worker process and we ended up here
# it's a fluke caused by the database instability
# we have to bail out and hope the service daemon survives
# since the default logger will also be affected by this error, we can't log anything either :(
if ( (getppid() > 1) || ($WORKER_PROCESS == 1) ) { exit(42); }
if ($DEBUG_MODE) {
print "EXCEPTION THROWN: ",$e->text,"\n";
print "ATTEMPTING TO RECONNECT...\n";
}
my $retry;
my $return_code = 0;
for($retry = 1; $retry <= $SAFE_MODE_RETRIES; $retry++) {
my $success = 0;
try {
$success = $worker->dbConnect();
} otherwise {
# actually, if we fail, we do nothing
};
if ($success) {
$return_code = 1;
last;
}
sleep $SAFE_MODE_DELAY;
}
if ($DEBUG_MODE && $return_code) {
print "DATABASE CONNECTION REESTABLISHED!\n";
} elsif ($DEBUG_MODE) {
print "DATABASE RECONNECTION ATTEMPTS FAILED!\n";
}
if ($return_code) {
$worker->logMsg(LOG_CRIT, "Exiting SAFE MODE; Reestablished connection to database after exception: ".$e->text);
$worker->errstr(undef);
redo MAIN_LOOP;
} else {
$worker->errstr("Unable to reconnect to ".$params->{dsn}.": ".$e->text);
$CLEAN_SHUTDOWN = 0;
last MAIN_LOOP;
}
};
} # end of MAIN_LOOP
# shutdown procedures
if ($CLEAN_SHUTDOWN) {
# if we're cleanly shutting down (we were told to via helios_params_tb)
# unregister from Helios database
# remove our PID
terminator();
} else {
# if we've had a db error, this will throw an exception when it tries to log to the database
# BUT BEFORE it does that, the error will have been logged to the local syslog daemon
$worker->logMsg(LOG_ERR,"$0 $worker_class HALTED on error: ".$worker->errstr);
}
if ($DEBUG_MODE) { print "$0 $worker_class HALTED!\n"; }
exit();
=head1 SUBROUTINES
=head1 PROCESS CONTROL FUNCTIONS
=head2 daemonize()
The daemonize() function is called to turn the Helios program into a daemon
servicing jobs of a particular class. It forks a new process, which disconnects
from the launching terminal.
Normally, daemonization would also include setting up signal handling, but the
daemonize() function isn't called in debug mode, so signal traps are actually
set up in the main program.
=cut
sub daemonize {
# prep system interaction stuff
# attempt to be a good daemon (unlike in the past)
chdir File::Spec->rootdir();
# don't think I need these w/close()ing them below,
# but I'll put them in for perlipc's sake anyway
open STDIN, '/dev/null';
open STDOUT, '>/dev/null';
open STDERR, '>/dev/null';
my $pid = fork;
# make sure fork was successful
unless ( defined($pid) ) {
die "Cannot daemonize: $!";
}
# we forked, but are we parent or child?
if ($pid) {
# we got a PID, so we're the parent ("shell-called process")
# print a nice message about daemonizing and exit
print "$worker_class daemon launched.\n";
exit();
}
# we got 0, so we're the "child", or the parent daemon
# we need to write our PID out to a file
# and then disconnect completely from the shell
write_pid_file($params->{pid_path}) or die($worker->errstr);
# Detach from the shell entirely by setting our own
# session and making our own process group
# as well as closing any standard open filehandles.
POSIX::setsid();
#[] I may have changed religions on this one
# close (STDIN);
# close (STDOUT);
# close (STDERR);
# set up signal handling in main process
}
=head2 double_clutch()
The double_clutch() function implements the WORKER_MAX_TTL functionality. If
the WORKER_MAX_TTL parameter is set for a service, the service daemon
periodically calls double_clutch() to check the workers and clean up any that
have run too long. The double_clutch() function waits a certain amount of time
(WORKER_MAX_TTL_WAIT_INTERVAL secs), then checks the amount of time each of the
running workers has been active. If a worker has been running longer than the
service's WORKER_MAX_TTL, the worker is killed (by sending it a SIGKILL signal).
=cut
sub double_clutch {
# BEGIN CODE Copyright (C) 2012 by Logical Helion, LLC.
# check each running worker to see if it has been running too long
# (too long: WORKER_MAX_TTL seconds + WORKER_MAX_TTL_WAIT_INTERVAL "fudge factor")
# END CODE Copyright (C) 2012 by Logical Helion, LLC.
foreach my $pid (keys %workers) {
# BEGIN CODE Copyright (C) 2012 by Logical Helion, LLC.
my $time_to_die = $workers{$pid} + $params->{WORKER_MAX_TTL} + $WORKER_MAX_TTL_WAIT_INTERVAL;
# END CODE Copyright (C) 2012 by Logical Helion, LLC.
if ( time() > $time_to_die ) {
kill 9, $pid;
delete $workers{$pid};
$worker->logMsg(LOG_ERR, "Killed process $pid (exceeded WORKER_MAX_TTL)");
}
}
return 1;
}
=head2 launch_worker()
#[] fix this POD - Helios::TS not TheSchwartz
The launch_worker() function launches a new worker process.
After the fork() from the main process, the new child process will call
launch_worker(). The launch_worker() function will instantiate a new
TheSchwartz object, set the object's B<ability> (TheSchwartz term) to the class
of the loaded service, and starts the worker on its way by calling the work()
method. If the OVERDRIVE run mode is enabled for the service, the
work_until_done() method is called instead.
=cut
sub launch_worker {
# just in case this would cause a problem in worker process
$SIG{CHLD} = 'DEFAULT';
$SIG{TERM} = 'DEFAULT';
# BEGIN CODE Copyright (C) 2012-3 by Logical Helion, LLC.
# [LH] [2013-09-21]: Added code to enable job prioritization features in Helios::TheSchwartz.
# [LH] [2013-10-04]: Virtual jobtypes: switched to using Helios::TS.
my Helios::TS $client = Helios::TS->new(
databases => $DATABASES_INFO,
prioritize => defined($params->{PRIORITIZE_JOBS}) ? $params->{PRIORITIZE_JOBS} : $DEFAULTS{PRIORITIZE_JOBS}
);
# END CODE Copyright (C) 2012-3 by Logical Helion, LLC.
$client->can_do($worker_class);
# BEGIN CODE Copyright (C) 2013 by Logical Helion, LLC.
# [LH] [2013-10-04]: Virtual jobtypes: Set up Helios::TS object with
# alt jobtypes and designate the "Active Worker Class" (the service class
# that is actually running the jobs).
if ($worker->getAltJobTypes) {
$client->set_active_worker_class($worker_class);
for ($worker->getAltJobTypes) {
$client->can_do($_);
}
}
# END CODE Copyright (C) 2013 by Logical Helion, LLC.
my $return;
if ( defined($params->{OVERDRIVE}) && $params->{OVERDRIVE} == 1 ) {
$return = $client->work_until_done();
} else {
$return = $client->work_once();
}
exit($return);
}
=head1 SIGNAL HANDLERS
=head2 reaper()
The reaper() function is responsible for cleaning up after dead child processes. It's called when
helios.pl receives a SIG_CHLD signal. The function reaps any children with waitpid(), removes the
children's PID from the $workers hash of running workers, and re-establishes itself as the signal
handler for the next SIG_CHLD signal.
=cut
sub reaper {
my $pid;
if ($DEBUG_MODE) { print "REAPING!\n"; }
while (($pid = waitpid(-1, &WNOHANG)) > 0) {
if ($pid == -1) {
if ($DEBUG_MODE) { print "REAPED IGNORING\n"; }
# no child waiting. Ignore it.
} elsif (WIFEXITED($?)) {
delete $workers{$pid};
if ($DEBUG_MODE) { print "REAPED Process $pid exited.\n"; }
if ($DEBUG_MODE) { print "REAPED $pid: $?\n"; }
} else {
if ($DEBUG_MODE) { print "REAPED False alarm on $pid.\n" }
}
}
if ($DEBUG_MODE) { print "FINISHED REAPING\n"; }
$SIG{CHLD} = \&reaper;
if ($DEBUG_MODE) { print "EXIT REAPER!\n"; }
}
=head2 terminator()
The terminator() function is responsible for shutting down a Helios service
instance when helios.pl receives a SIGTERM signal. The shutdown process
performs several steps:
=over 4
=item *
Sets HALT in the Helios config params for the loaded service, so the
worker processes know to exit at the end of their current job
=item *
Sleeps a certain amount of time (zero_launch_interval x 2 secs) to
allow worker processes to complete the jobs they are processing and exit
=item *
Reaps any ended worker processes, and forcably kills any that are still
running. After zero_launch_interval seconds, workers for a particular service
should have exited if HALT is set. If you use this method of shutting down
Helios services and repeatedly see a particular service's workers not exiting
properly, increase the value of the zero_launch_interval configuration
parameter for that service in your helios.ini or Helios::Panoptes Ctrl Panel.
=item *
Clears the HALT parameter set earlier, since all of the workers are now
shut down (one way or the other).
=item *
Removes the service's PID file written in the pid_path set in
helios.ini
=item *
Unregister's with the Helios collective's registry table in the Helios
database
=item *
Issues a final log message indicating the loaded service has shut down
on the current host
=back
Once all these steps are complete, the helios.pl program exits.
=cut
sub terminator {
my $TERM = 0;
# did we receive a SIGTERM, or was the HALT config param set?
if ( defined($params->{HALT}) ) {
$worker->logMsg(LOG_NOTICE, "HALTING $CLASS on host ".$worker->getHostname());
} else {
# tell the workers they need exit
$TERM = 1;
$worker->logMsg(LOG_NOTICE, "Received TERM signal; HALTING $CLASS on host ".$worker->getHostname());
$worker->logMsg(LOG_NOTICE, "Setting HALT for $CLASS on host ".$worker->getHostname());
set_halt();
}
# sleep long enough for the workers to shutdown on their own
$SIG{CHLD} = 'DEFAULT';
sleep $ZERO_LAUNCH_INTERVAL;
sleep $ZERO_LAUNCH_INTERVAL;
# reap any processes that ended while we were sleeping
$worker->logMsg(LOG_NOTICE, "Reaping $CLASS workers on host ".$worker->getHostname());
reaper();
# kill any workers still running (they've had plenty of time to end on their own)
foreach my $pid (keys %workers) {
if ( kill 0 => $pid ) {
# it's still alive, kill it
kill 9, $pid;
$worker->logMsg(LOG_ERR, "Killed process $pid (shutdown $CLASS instance)");
}
delete $workers{$pid};
}
# workers are all taken care of, shutdown ourself
# only clear the HALT if we received a SIGTERM (we set the halt ourselves)
if ($TERM) {
clear_halt();
$worker->logMsg(LOG_NOTICE, "Cleared HALT for $CLASS on host ".$worker->getHostname());
}
clean_shutdown();
$worker->logMsg(LOG_NOTICE, "$CLASS on host ".$worker->getHostname().' HALTED.');
exit(1);
}
# [LH] 2013-08-04: Added 2nd paragraph to write_pid_file() documentation to
# help explain how new running_process_check()/write_pid_file() works.
=head1 PID FILE FUNCTIONS
=head2 write_pid_file($pid_path)
Writes a PID file to a location (defaults to /var/run/helios) to track which daemons are
running. The file will be named after the service class running, all lowercase, with colons
replaced by underscores. For example, the PID file for a service class named
'SearchIndex::LoadTestWorker' will be named "searchindex__loadtestworker.pid". To change the
location where the PID file is created, set the pid_path option in helios.ini.
The running_process_check() function should be run before this function to
exclusively lock the PID file.
=cut
sub write_pid_file {
# BEGIN CODE Copyright (C) 2013 Logical Helion, LLC.
# [LH] 2013-08-04: Rewrote write_pid_file() to take advantage of the
# exclusive lock on the pidfile created by running_process_check().
# [RT81914]
# Rewind the PID file handle opened by running_process_check(),
# write our own PID in the file, and
# signal to the calling routine it's OK to finish start up.
seek($PID_FILE_H, 0, 0) or do {
$worker->errstr("Cannot rewind $PID_FILE: ".$!);
close($PID_FILE_H);
return 0;
};
print $PID_FILE_H $$,"\n";
truncate($PID_FILE_H, tell($PID_FILE_H)) or do {
$worker->errstr("Cannot truncate $PID_FILE after writing PID: ".$!);
close($PID_FILE_H);
return 0;
};
close($PID_FILE_H);
# END CODE Copyright (C) 2013 Logical Helion, LLC
return 1;
}
=head2 remove_pid_file($pid_file)
During a clean shutdown, remove_pid_file() is called to delete the PID file
associated with the service daemon.
=cut
sub remove_pid_file {
my $pid_file = shift;
unless ( defined($pid_file) ) {
$pid_file = $PID_FILE;
}
unlink $pid_file or do { $worker->errstr("Cannot remove PID file $pid_file: ".$!); return undef; };
return 1;
}
# [LH] 2013-08-04: Added 2nd paragraph below to explain how new running_process_check() works.
=head2 running_process_check($pid_path)
Given the pid_path, check to see if a $pid_file for the loaded service class exists and, if it does,
check to see if that process is still running. If the file doesn't exist or it does but isn't
running, this function returns 0. If the process is still running, record the error and return the
running process's pid to signal that service startup should halt.
If the PID file does not exist, running_process_check() creates it. The PID
file is then exclusively locked to prevent other daemons for the same service
from writing to it. This helps ensure multiple Helios daemons for same
service will not start up at the same time.
=cut
sub running_process_check {
my $pid_path = shift;
# Determine where the PID file should be
unless (defined($pid_path) ) { $pid_path = $DEFAULTS{PID_PATH}; }
# Determine PID filename
my $filename = lc($worker_class);
$filename =~ s/\:/\_/g;
$PID_FILE = File::Spec->catfile($pid_path, $filename.'.pid');
if ($DEBUG_MODE) { print "Checking $PID_FILE\n"; }
# check if this file exists
# if it does, check if that process is still running
# bail if it is
# BEGIN CODE Copyright (C) 2013 Logical Helion, LLC.
# [LH] 2013-08-04: Rewrote this piece of running_process_check() to put
# an exclusive lock on the pidfile (also creating it if it doesn't exist).
# Also switched to using the relatively portable Perl kill() instead of
# grep-ing shell output to determine whether the process in the pidfile
# is still running. [RT81914]
sysopen($PID_FILE_H, $PID_FILE, O_RDWR | O_CREAT) or do {
$worker->errstr("Cannot open $PID_FILE: ".$!);
return 1;
};
flock($PID_FILE_H, LOCK_EX | LOCK_NB) or do {
$worker->errstr("Cannot lock $PID_FILE (another daemon already starting for this service?): ".$!);
close($PID_FILE_H);
return 1;
};
my $pid_in_file = <$PID_FILE_H>;
chomp $pid_in_file if defined($pid_in_file);
# if there's an actual, valid pid in the file,
# check to see if that process is still running
if ( defined($pid_in_file) && $pid_in_file !~ /\D/ && $pid_in_file > 1) {
my $proc_count = kill 0, $pid_in_file;
if ($proc_count > 0) {
$worker->errstr($worker->getJobType()." service daemon already running (process ".$pid_in_file.").");
close($PID_FILE_H);
return $pid_in_file;
}
}
# END CODE Copyright (C) 2013 Logical Helion, LLC.
return 0;
}
=head1 OTHER FUNCTIONS
=head2 clean_shutdown()
The clean_shutdown function is called when helios.pl is intentionally shutdown
(setting a HALT parameter in the Helios::Panoptes Ctrl Panel or sending the
helios.pl process a TERM signal). It removes the PID file created on
startup and unregisters the service instance from the collective database.
=cut
sub clean_shutdown {
remove_pid_file() or $worker->logMsg(LOG_CRIT, $worker->errstr);
unregister() or $worker->logMsg(LOG_CRIT, $worker->errstr);
return 1;
}
=head2 register()
The register() function records information about the currently running worker daemon in the
database. The register() function is designed to be run every $REGISTRATION_INTERVAL seconds.
That way, if a service daemon dies off unexpectedly (without calling unregister()), it can be
determined that something has happened to the daemon and it possibly needs to be restarted.
(In reality, register() and unregister() are only necessary to provide a display for Panoptes,
to more easily assess system status and facilitate the HALTing of service daemons or HOLDing of
job processing.)
=cut
sub register {
try {
my $dbh = $worker->dbConnect();
$dbh->do("DELETE FROM helios_worker_registry_tb WHERE worker_class = ? AND host = ?", undef,
$worker->getJobType(), $worker->getHostname) or die;
$dbh->do("INSERT INTO helios_worker_registry_tb (register_time, start_time, worker_class, worker_version, host, process_id) VALUES (?,?,?,?,?,?)", undef,
time(), $START_TIME, $worker->getJobType(), $worker->VERSION, $worker->getHostname, $$) or die;
} otherwise {
throw Helios::Error::DatabaseError($DBI::errstr);
};
return 1;
}
=head2 unregister()
The unregister() function removes any record of the currently running daemon from the database.
It is called whenever there is a clean shutdown.
=cut
sub unregister {
try {
my $dbh = $worker->dbConnect();
$dbh->do("DELETE FROM helios_worker_registry_tb WHERE worker_class = ? AND host = ?", undef,
$worker->getJobType(), $worker->getHostname) or die;
$dbh->disconnect();
} otherwise {
throw Helios::Error::DatabaseError($DBI::errstr);
};
return 1;
}
=head2 set_halt()
Set a HALT parameter for the currently loaded service on the current host to
signal to all of the worker processes that they need to exit. This function is
used by the terminator() function to safely signal to workers they need to exit
when the current job is completed.
=cut
sub set_halt {
try {
if ($DEBUG_MODE) {
print "Attempting to set HALT for ",$worker->getJobType()," on ",$worker->getHostname(),"\n";
}
my $dbh = $worker->dbConnect();
$dbh->do("INSERT INTO helios_params_tb (worker_class, host, param, value) VALUES (?,?,?,?)", undef,
$worker->getJobType(), $worker->getHostname, 'HALT', '1');
} otherwise {
throw Helios::Error::DatabaseError($DBI::errstr);
};
return 1;
}
=head2 clear_halt()
If the --clear-halt option is specified on the command line, clear_halt() is called to attempt to
clear the HALT parameter in the HELIOS_PARAMS_TB. For safety reasons, it only clears a HALT for
the loaded service class AND the specific host helios.pl is running on; it will not clear a global
HALT parameter (where the host is specified as '*').
=cut
sub clear_halt {
try {
if ($DEBUG_MODE) {
print "Attempting to delete HALT for ",$worker->getJobType()," on ",$worker->getHostname(),"\n";
}
my $dbh = $worker->dbConnect();
$dbh->do("DELETE FROM helios_params_tb WHERE worker_class = ? AND host = ? AND param = ?", undef,
$worker->getJobType(), $worker->getHostname, 'HALT');
} otherwise {
throw Helios::Error::DatabaseError($DBI::errstr);
};
return 1;
}
# BEGIN CODE Copyright (C) 2012 by Logical Helion, LLC.
# [LH] 2012-10-12: Added this method to more securely load module code at runtime.
sub require_module {
my $service_class = shift;
if ( $service_class !~ /^[A-Za-z]([A-Za-z0-9_\-]|:{2})*[A-Za-z0-9_\-]$/ ) {
print STDERR "Sorry, requested name is invalid: $service_class\n";
exit(1);
}
unless ( $service_class->can('new') ) {
eval {
my $class_file = $service_class;
$class_file .= '.pm';
$class_file =~ s/::/\//g;
require $class_file;
1;
} or do {
my $E = $@;
print STDERR "Perl module $service_class could not be loaded: $E\n";
exit(1);
};
}
unless ($service_class->isa('Helios::Service')) {
print STDERR "$service_class is not a subclass of Helios::Service.\n";
exit(1);
}
return 1;
}
# END CODE Copyright (C) 2012 by Logical Helion, LLC.
# BEGIN CODE Copyright (C) 2012 by Logical Helion, LLC.
# [LH] [2013-09-21]: Code to implement WORKER_LAUNCH_PATTERN feature.
=head2 workers_to_launch($waiting_jobs, $running_workers, $max_workers)
=cut
sub workers_to_launch {
my $waiting_jobs = shift;
my $running_workers = shift;
my $max_workers = shift;
my $wlp = defined($params->{WORKER_LAUNCH_PATTERN}) ? $params->{WORKER_LAUNCH_PATTERN} : $DEFAULTS{WORKER_LAUNCH_PATTERN};
# how many workers are we able to launch?
my $available_workers = $max_workers - $running_workers;
# if we cannot launch any more workers, then we cannot launch any workers
if ( $available_workers <= 0 ) {
return 0;
}
# if the waiting jobs >= max_workers, forget all the launcher patterns and blitz
if ( $waiting_jobs >= $max_workers ) {
return $available_workers;
}
# ok, we might be able to launch some workers
SWITCH: {
# opportunistic
# for big, fast collective databases with lots of short running jobs
if ( $wlp eq 'oppt') {
# launch a worker for each waiting job,
# or all available workers
# whichever is less
if ( $available_workers < $waiting_jobs ) {
return $available_workers;
} else {
return $waiting_jobs;
}
}
# progressive
# for fast collectives that need to launch workers more slowly
# e.g. an app with a slow or limited shared resource
# this will be the the Helios 3.x default
if ( $wlp eq 'prog') {
# launch the difference between running workers and waiting jobs
# e.g. 14 waiting jobs, 10 workers running, launch 4
# if waiting jobs exceed available workers, just launch available workers
my $diff = $waiting_jobs - $running_workers;
if ( $diff <= 0 ) { $diff = 1; }
if ( $available_workers < $diff ) {
return $available_workers;
} else {
return $diff;
}
}
# conservative
# for slow collective databases, or apps that only run a
# smaller number of jobs, or longer running jobs
# this was the Helios 2.x default
if ( $wlp eq 'cons') {
return 1;
}
# the default is actually 'conservative'
return 1;
}
}
# END CODE Copyright (C) 2012 by Logical Helion, LLC.
=head1 SEE ALSO
L<Helios>, L<Helios::Service>
=head1 AUTHOR
Andrew Johnson, E<lt>lajandy at cpan dotorgE<gt>
=head1 COPYRIGHT AND LICENSE
Copyright (C) 2007-9 by CEB Toolbox, Inc.
Portions of this software, where noted, are
Copyright (C) 2012-2013 by Logical Helion, LLC.
This program is free software; you can redistribute it and/or modify
it under the same terms as Perl itself, either Perl version 5.8.0 or,
at your option, any later version of Perl 5 you may have available.
=head1 WARRANTY
This software comes with no warranty of any kind.
=cut