135/netdisco
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

rename pollers to tasks

Browse Source
This commit is contained in:
Oliver Gorwits
2014-08-10 14:45:22 +01:00
parent 44b50f413f
commit a13ed25f6a
8 changed files with 31 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
Netdisco/lib/App/Netdisco/Configuration.pm
View File

@@ -51,11 +51,11 @@ setting('workers')->{queue} ||= 'PostgreSQL';
if (exists setting('workers')->{interactives}
or exists setting('workers')->{pollers}) {
setting('workers')->{pollers} = (
setting('workers')->{tasks} =
(setting('workers')->{pollers} || 0)
.' + '. (setting('workers')->{interactives} || 0)
) if setting('workers')->{interactives};
+ (setting('workers')->{interactives} || 0);
delete setting('workers')->{pollers};
delete setting('workers')->{interactives};
}

2
Netdisco/lib/App/Netdisco/Daemon/Worker/Manager.pm
View File

@@ -39,7 +39,7 @@ sub worker_body {
if setting('workers')->{'no_manager'};
my $num_slots =
MCE::Util::_parse_max_workers( setting('workers')->{pollers} )
MCE::Util::_parse_max_workers( setting('workers')->{tasks} )
- $self->{queue}->pending();
while (1) {

4
Netdisco/lib/App/Netdisco/Daemon/Worker/Scheduler.pm
View File

@@ -15,7 +15,7 @@ sub worker_begin {
my $wid = $self->wid;
return debug "mgr ($wid): no need for scheduler... skip begin"
unless setting('scheduler');
unless setting('schedule');
debug "entering Scheduler ($wid) worker_begin()";
@@ -39,7 +39,7 @@ sub worker_body {
my $wid = $self->wid;
return debug "mgr ($wid): no need for scheduler... quitting"
unless setting('scheduler');
unless setting('schedule');
while (1) {
# sleep until some point in the next minute

13
Netdisco/lib/App/Netdisco/Manual/Configuration.pod
View File

@@ -879,14 +879,17 @@ or remove any entries.
Value: Settings Tree. Default:
workers:
pollers: 'AUTO * 2'
tasks: 'AUTO * 2'
sleep_time: 1
Control the activity of the backend daemon with this configuration setting.
C<pollers> sets how many workers are started for interactive jobs (port
C<tasks> sets how many worker processes are started for interactive jobs (port
control) and polling jobs (discover, macsuck, arpnip) on this node. Other
nodes can have different settings. "C<AUTO>" is the number of CPU cores.
nodes can have different settings.
"C<AUTO>" is the number of CPU cores. Set C<tasks> to "C<0>" to disable all
workers (which allows you to have a scheduler-only node).
C<sleep_time> is the number of seconds between polling the database to find
new jobs. This is a balance between responsiveness and database load.
@@ -918,9 +921,9 @@ If set, then this node's backend daemon will schedule polling jobs (discover,
macsuck, arpnip, etc) in the central database. It's fine to have multiple
nodes scheduling work for redundancy (but make sure they all have good NTP).
Note that this is independent of the Pollers configured in C<workers>. It's
Note that this is independent of the Tasks configured in C<workers>. It's
okay to have this node schedule schedule but not do any of the polling
itself (C<pollers: 0>).
itself (C<tasks: 0>).
Work can be scheduled using C<cron> style notation, or a simple weekday and
hour fields (which accept same types as C<cron> notation). For example:

9
Netdisco/lib/App/Netdisco/Manual/ReleaseNotes.pod
View File

@@ -48,6 +48,15 @@ web interface. We advise that you back up the database prior to upgrade:
=head2 General Notices
The backend polling daemon has been rewritten and as a result your
configuration can be simplified. Some keys have also been renamed. Our advice
is to remove (or comment out) the complete C<workers> configuration which
enables auto-tuning. If you do wish to control the number of worker
processes, follow this pattern:
workers:
tasks: 'AUTO * 2' # this is the default, twice the number of CPUs
The configuration item C<reports> is now a list (used to be a dictionary).
Each item in the list must have a C<tag> entry which was previously the
dictionary key. For example, now use: