Beefy Boxes and Bandwidth Generously Provided by pair Networks
Think about Loose Coupling
 
PerlMonks  

Mojo: non-blocking API calls

by Beatnik (Parson)
on Nov 27, 2016 at 21:30 UTC ( #1176664=perlquestion: print w/replies, xml ) Need Help??

Beatnik has asked for the wisdom of the Perl Monks concerning the following question:

Hey All, I'm continuing my toy project and I'm looking at the responsiveness. Part of the functionality involves external API calls. I've developed the Perl modules that do the external calls (through LWP at this point). My application loads data locally from a database, allows the user to modify the database entry but also should fire off a 'save' call to the APIs. There is some querying from the API to the local database. Obviously, this is expensive so I'm trying to use a non-blocking call. In addition, since users can use the applications, some kind of checks will have to be done to reload application data locally as clean as possible. I've been looking at Mojo::IOLoop::Delay to implement this.

My question is: do I rewrite / optimize the API library to add non-blocking support (swapping out LWP::UserAgent)? Is Mojo::IOLoop::Delay the right approach for this purpose? Are there any recommended approaches for the synchronizing from the API to the database?


Greetz
Beatnik
... I'm belgian but I don't play one on TV.

Replies are listed 'Best First'.
Re: Mojo: non-blocking API calls
by Anonymous Monk on Nov 28, 2016 at 02:13 UTC
Re: Mojo: non-blocking API calls
by stevieb (Abbot) on Nov 27, 2016 at 21:49 UTC

    If feasible, please try to show some of the relevant code snips you have now, perhaps with comments in the areas you feel that could use some performance gains.

      No specific code for the non-blocking right now. I'm currently only loading from the database but the release candidate will need to work the application. Each API call can take a few seconds. With bulk changes on the data, I'm trying to limit the wait for the user. I was considering Minion to run the synchronization process as a background job but that won't update the application immediately. My API module is Net::Cisco::ACS, which is based on LWP::UserAgent. As far as I can tell from looking at Mojo::UserAgent, I should not have any issue swapping out the code. My only concern is where I would need to add the IOLoop block? Would my module method need something extra? I'll need to do some more reading on this, I guess :)


      Greetz
      Beatnik
      ... I'm belgian but I don't play one on TV.
        "I'll need to do some more reading on this, I guess"

        Quite probable. From what you're describing, it's a shot in the dark unless one of the monks has experience in exactly what you're dealing with.

        After reading more, and with some testing, you should be able to provide code snips that show the issue in action. I can think of a few things off the bat, but I'm afraid they'll be irrelevant and slant this thread sideways for no reason.

        For instance... in one of my apps recently using Dancer2, I had to sway from their db plugin so that I could run my db inserts/updates inside of an event outside the main application, so that there was no holdup. That won't work in all cases though, so keep us apprised on what you find when you can get a bit more specific.

      Pseudo code would be something like:
      sub save { my $self = shift; my $acs = Net::Cisco::ACS->new("hostname"=>"localhost"); my $user = Net::Cisco::ACS::User->new("username"=>"foobar"); $acs->create($user); # Blocking code # Calling create will take a few seconds } sub list { my $self = shift; my $acs = Net::Cisco::ACS->new("hostname" => "localhost"); my %users = $acs->users; # Blocking code # Calling users will take a few seconds }
      I don't need to call the API at that specific time. My original approach was to flag the record in my local database as 'created' or 'modified' and run a background process to make sure the local database synchronizes all flagged records with the application using the API. The list method would then only read from the database and make adjustments in the database.


      Greetz
      Beatnik
      ... I'm belgian but I don't play one on TV.

        Perhaps something as simple as forking is all you need. Here's an example that kind of simulates what I think you're trying to do. Essentially, we create an API object with a couple of methods. Instead of calling them directly, we pass the object and the sub we want to call on it to a forking routine from within a local sub (save(), users()), and then let it work in the background while the main script carries on.

        At the end of the script, we wait until all forked processes have completed before we exit.

        use warnings; use strict; package Thing; { sub new { return bless {}, shift; } sub users { print "in ". __PACKAGE__ . " execing users()\n"; sleep 2; # simulate long API call print "done API users()\n"; } sub save { print "in ". __PACKAGE__ . " execing save()\n"; sleep 2; # simulate long API call print "done API save()\n"; } } package main; use Parallel::ForkManager; my $max_forks = 10; my $fork = new Parallel::ForkManager($max_forks); my $api = Thing->new; print "doing stuff in main...\n"; print "calling main::save()...\n"; save($api); print "continuing in main, save() API call isn't blocking\n"; print "calling main::users()...\n"; users($api); print "finished in main, waiting for forks to finish\n"; $fork->wait_all_children; sub save { my $api = shift; fork_api_call($api, 'save'); print "forked API save...\n"; } sub users { my $api = shift; fork_api_call($api, 'users'); print "forked API users...\n"; } sub fork_api_call { my ($api, $sub) = @_; for (0..$max_forks){ $fork->start and last; $api->$sub(); $fork->finish; } }

        Output:

        doing stuff in main... calling main::save()... forked API save... continuing in main, save() API call isn't blocking calling main::users()... in Thing execing save() forked API users... finished in main, waiting for forks to finish in Thing execing users() done API save() done API users()

Log In?
Username:
Password:

What's my password?
Create A New User
Node Status?
node history
Node Type: perlquestion [id://1176664]
Approved by marto
Front-paged by marto
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others perusing the Monastery: (4)
As of 2019-07-20 10:16 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    No recent polls found

    Notices?