Pages

Wednesday, November 6, 2013

Jobs, and how they save your life

No. I don't mean your job. You are reading this page because you have a job and you need your life to be easier, or you don't have a job, and cats got boring. What I do know is Powershell, and it's ability to thread workloads, has made me keep my current job, as well as other important things like sanity, clean underpants, hair, carpel-tunnel free hands. You get the idea. It's also a topic that when I was starting out learning about jobs, I had trouble finding many good, easy to use examples of how I could apply them for what I need.

There are many types of jobs in Powershell, and just as many different ways to execute them. However, for the most part, you either have a script you wish you could thread, or you have the requirement to thread a lot of things instead of running them sequentially. A few instances this might be used in is, for example, starting or stopping services on hundreds of servers, copying a file to a remote file system on each server, creating/nesting groups based off of a document in AD, grabbing log files from many different hosts, or whatever it is you may be doing. I hope the valuable application of this has already become apparant.

Lets get something clear first. Start-Job is expensive. It also starts a seperate instance of Powershell. This brings with it two unique challenges:
  1. You are going to catch your computer or server on fire (maybe)
  2. You can't pass variables from the host shell so easily
Number 1 is because Start-Job creates a new instance of powershell, which needs resources. Personally, I have gotten away with running about 150 Powershell instances concurrently, which required about 12GB of available memory. Number 2 is valid. Since it is spinning up a fresh clean instance of Powershell, it won't have any of your session variables stored, so you need to pass them to the job. This would be the function of -ArgumentList. We will go over this in this post.

Pancake:
Start-Job -Name JobNameWhatever -Scriptblock {..} -ArgumentList $var,$var2

Easy enough right? This is the basic breakdown of Start-Job, and how 99% of the worlds population will be using it. You can start a job based on a remote script as well, which is similar to Invoke-Expression (We won't cover that here), you can use a very large and complicated script block (what you will most likely be doing), or you can just enter any plain old boring Powershell command, which you will put into a script block anyways. So let's take another example of actually using Start-Job in a loop. This will basically serve all of the needs I think you need to use it for.

$Names = Get-Content C:\PowershellPancakes\AuntJemima.txt
foreach ($Name in $Names)
    {
    $MapleSyrup =
        {
        Param ($name)
        New-Item -Name $Name -ItemType Directory -Path C:\PowershellPancakes
        }

Start-Job -Name $Name -Scriptblock $MapleSyrup -ArgumentList $Name
    }

This is just a simple script in a loop, where we are creating some delicious folders based off the names in the text file. You can open the loop however you want, in any way you want. This is the basic function. So imagine your $MapleSyrup scriptblock. You can do whatever you want in this script block. You can even take an existing script that takes manual input, put it into a script-block like so, and start a job based off of it. I have been doing this increasingly so, when someone asks me, can we do this, like, all at once? Why yes, client A, we can do this all at once. To maintain your full head of hair I suggest you put the job name to a variable you can identify. That way if you are running several hundred jobs you can skim through a list with Get-Job.

I want to go into detail about the following bit, because this may cause you a lot of headaches when you get started with jobs. So here it is:

Param ($name)

Were you expecting more? This just needs to be noted. This is where your parameters will get transferred into your scriptblock, which need to be used in -ArgumentList by Start-Job. If you don't, the value of $name will be null when the job starts, which could create unforseen problems, as well as just not work. Another note, variables you create inside script blocks don't need to be added to -ArgumentList or Param. They are contained. As long as you add the external variables being passed from outside of the scriptblock through Param and -ArgumentList, you are okay. Also remember, seperate params by a comma, like so,

Param ($name,$plants,$syrup)


Or, personally,

Param ($name,
       $syrup,
       $plants)


You have just one line you need to run?

Start-Job -Name Blueberries -Scriptblock {robocopy \\serverB\1 \\serverB\2}

What's nice about starting jobs is since they use another instance, you're free to continue using the parent shell for whatever you want. Some of you may have run into the situation where if you were running a script in the ISE and wanted to do something else on another script in the same ISE, you couldn't. Well since Start-Job creates its own instance of Powershell, you don't have to worry about this.

So you've started a job and it's just...running...somewhere...

Get-Job 

Though no, you can't exactly expand the operating job, and you will need to have some confidence in your own script. In another post we will go over how you can add some error handling to pass things back to your parent shell, or to a log file using different commands, including Try and Catch, which we touched on earlier.

But I have a job I need to run, through 400 objects in my CSV file that i'm looping through. I only have 4gb of available memory! Oh what do I do? First I will refer you to a previous posting, which goes very closely with this one. That would be What Do Until loops can do for you. They are connected directly. But how? Let's take our earlier example, and put some job control on it.

$Names = Get-Content C:\PowershellPancakes\AuntJemima.txt
foreach ($Name in $Names)
    {
    $MapleSyrup =
        {
        Param ($name)
        New-Item -Name $Name -ItemType Directory -Path C:\PowershellPancakes
        }

Start-Job -Name $Name $MapleSyrup -ArgumentList $Name
DO {
    Start-Sleep -Seconds 5
    #Give it some time
    Remove-Job -State Completed
    } Until (@(Get-Job).Count -le 5)
    }

What just happened? Well, we just limited the number of active jobs at any given time. I probably don't need to tell you why you want to do this. Especially if you are working with just huge bits of data and lots of jobs need to be created to do it. Now you can create jobs that create sub-jobs, but that's just a bit unnecessary in my personal opinion, unless each job has a lot of items it needs to complete, that would be better suited by threading those as well. Just remember your hardware limit so you don't do anything too crazy and bring everything to a halt. Also consider your network/storage/hardware or whatever it is you will be working with. Let's not be reckless here.

I think that basically covers it for Jobs. An absolutely essential function of Powershell that I think could sometimes be misunderstood. Mess around with the examples in this post and you should get a very good feel for how to incorporate these Pancakes into your own recipes. 



No comments: