Restarting a service in powershell vs restarting as myself yields different results - powershell

Context: I want to deploy some software automatically and restart the Apache2.4 service as part of that deployment.
I've been racking my brains trying to figure out why restarting a service via powershell is different from restarting it as myself via services.msc.
Basically I have a local site I'm browsing to. After deploying the files for the site, I need to restart the Apache2.4 service.
Scenario 1: If I restart the service myself (by going to services.msc and right clicking Restart) - then I can browse to my site.
Scenario 2: If I restart the service via powershell (via a batch file that calls a powershell script that runs the command Restart-Service -DisplayName "Apache2.4") then I can't browse to my site. No matter how many times I do it through powershell, it won't let me load the site.
In both scenario 1 and 2 above the service DOES restart. I have confirmed so via the event viewer. That's why I am so confused by what is happening. Has anyone come across this?

Related

How do I automatically load powershell profiles with Jenkins pipeline when running Jenkins as a service?

First off, I didn't have this issue until setting up my agent to run as a windows service.
My company has custom cmdlets we have built that are part of the default profile that is loaded when running powershell. I am using Jenkins to execute a batchfile that iterates a command over a series of machines. After settings up Jenkins to be a service, it no longer has access to those cmdlets leading me to believe the profile isn't being loaded. If I load the profile manually by running the profile script, it only seems to work on the first machine.
When setting up Jenkins as a service, I configured it to be the same user that I would manually run these scripts as if I were to login to the computer. I have verified it is using the proper user with $env:UserName.
I am at a loss as to why setting up jenkins as a windows service broke this. I could revert to using the command line to connect to Jenkins, but that doesn't always connect post server maintenance or after a power outage.
Did I configure something wrong or is there a way to load profiles instead of jenkins always running -NoProfile?
Update - I noticed when running $PROFILE it was set to a default profile location that did not exist. It seems when opening powershell manually on the machine it loads the AllUsersCurrentHost profile but this doesn't happen when using powershell from Jenkins when running as a service. I created the file location where it said it was using the profile and copied the default profile there and it works. I am still not sure why the behavior differs, but at least I found a solution.

Unable to start service in window 10 by using NSSM

I have create a small script file to test.
This my script.bat file.
sc create myService binpath= C:\Users\Admin\Desktop\test.bat start= auto
This is my test.bat file.
echo "Welcome to Wizard"
Problem Statement
I am unable to start the service from control panel Service section.
I get following error.
[SC] StartService FAILED 1053:
The service did not respond to the start or control request in a timely fashion.
That is why I am using nssm.
NOW what happening is that when I run following command on powershell
.\nssm install myService, I dialogue box appears. I give it the path of my script file and click on install service.
After successfull installation of service. I go to control panel -> Service -> click on start against myService but it get paused and following dialog box appears
How can I fix this?
Is there anyother way to do it without doing manual steps and not using third party tool.
I am doing all this on window 10. Do I need any server to perform this task?
NOTE: I cannot use Always up or window scheduler in my case.
The NSSM behaviour is caused by the script terminating almost instantly. Try the following script:
echo Hello World
pause
This should allow the service to start, but you will not necessarily see a console window. Even if you tick 'allow service to interact with desktop', it will not be your desktop that it interacts with!
Windows implements 'session zero isolation' as a security feature, and this essentially prevents services interacting with end user desktops.
In terms of a solution, it's possible to write Windows 'service' applications fairly simply using Visual Studio. It's outside my area of expertise, but based on the Windows applications I'm familiar with, you would generally have a user-mode application running to provide desktop interaction. The user-mode application can interact with services hosted by the service application.
Probably this is resolved by now, but in case it helps anyone, what saved the day for me was checking again my input in the arguments field in nssm. I had an extra "-" which created the error. To edit my service, I went via nssm edit <servicename>
I would also add on the fix that worked for me. I added "" (quotes) in the argument path and that solved the issue for me.

Jenkins windows slave service does not interact with desktop

I have followed this guide to install a jenkins slave on windows 8 as a service:
https://wiki.jenkins-ci.org/display/JENKINS/Installing+Jenkins+as+a+Windows+service#InstallingJenkinsasaWindowsservice-InstallSlaveasaWindowsservice%28require.NET2.0framework%29
I need to run a job that interact with the desktop (run an application that opens a browser etc.). So after I have installed the slave as a service (running jnlp downloaded from the master) I have changed the service "Log on" to "Allow to interact with display".
For some reason its only possible to enable this for the "Local System account" even though its recommended to run the service as a specified user, eg. jenkins.
But nothing happens when I execute the job, the browser is not opened. If I instead stop the service and just launch the slave through the jnlp file the job runs fine - the browser is opened.
Anybody had any luck interacting with the desktop when running a jenkins windows slave as a service?
Services run since Vista in Session 0 and the first user is now in Session 1. So you can't interact any longer. This is called Session 0 Isolation.
Microsoft explains this here and here. You have to use 2nd Program which uses IPC to communicate to the Service.
I had lots of issues running Jenkins in Windows using the service.
Instead I now disable the service and run it from CMD.
So open CMD.
cd C:\Program Files (x86)\Jenkins
java -Xrs -Xmx256m -Dhudson.lifecycle=hudson.lifecycle.WindowsServiceLifecycle -jar
jenkins.war --httpPort=9091
To resolve it, first create Windows auto-logon as I explain here:
https://serverfault.com/questions/269832/windows-server-2008-automatic-user-logon-on-power-on/606130#606130
Then create a startup batch for Jenkins agent (place it in Jenkins directory). This will launch agent console on desktop, and should allow Jenkins to interact with Windows GUI:
java -jar slave.jar -jnlpUrl http://{Your Jenkins Server}:8080/computer/{Your Jenkins Node}/slave-agent.jnlp
(slave.jar can be download from http://{Your Jenkins Server}:8080/jnlpJars/slave.jar)
EDIT :
If you're getting black screenshots (when using Selenium or Sikuli, for example), create a batch file that disconnects Remote Desktop, instead of closing the RDP session with the regular X button:
%windir%\system32\tscon.exe %SESSIONNAME% /dest:console
Consider running the Java slave server directly at startup and then using something to monitor and restart should the server go down (e.g., Kiwi Restarter).
Please check the services (# TestNode) make sure the "Interactive Services Detection" service is STARTED, by default the startup type is set to Manual, you may like to set it to automatic as well.
After service started, when you run your test in the Test Node, you will see something like the below:
Click on it and choose view the message
You will see the activities happen there. Hope this helps :D
Note: If login with other account and cannot view the Interative Services Detection prompt, restart the service again.
My Jenkins Service runs as user "jenkins" and all I did was to create Desktop folders in: C:\Windows\system32\config\systemprofile\desktop and if 64 bit Windows also in C:\Windows\SysWOW64\config\systemprofile\desktop - then it runs perfectly.
Make sure that Desktop folders are created as such:
%WINDOWS%/System32/config/systemprofile/Desktop
%WINDOWS%/SystemWOW64/config/systemprofile/Desktop
Presence of those can sometimes be mandatory while running some Java software as a Service.

Deployment not in a domain - psexec.exe or powershell remoting

I am working on an automated deployment process for a web application. The deployment will need to:
Deploy DB changes to database using sqlpackage.exe
Deploy reporting services reports to the reports server using the web service
Deploy web app to web server(s)
Deploy fonts for reports
among other things
The first two are reasonably straightforward to run from the web server, as the web service and db are contactable, and the tools to deploy run over the network.
From reading it appears that powershell remoting should be the way to go, and internally this would not be a problem. However when deploying to production, this will be being carried out in a datacentre, where the machines (2web, 1db) are not on a domain at all. I'd like to come up with a generic process that can run both internally and externally with the appropriate configuration. Powershell remoting, with machines not in a domain appears to require a fair bit of configuration using https etc., as NT credentials can't be validated.
Should I battle out configuring powershell remoting, or would configuring this to just use psexec to execute a powershell script directly on the remote machien, copying the deployment artifacts to a drop folder on the remote machine be the best way to go?
psexec seems to "just work". It appears powershell remoting comes with a lot more pain.
Why not use psexec then? You can restrict it's role to just getting you on to the remote machine, and not let it infect your scripts. I have not attempted to get ps remoting working on a non-domain, but it general I have found it to be fairly high effort to get going. Psexec, as you say, can often be simpler.
Excuse the peddling, but the open source framework I helped build called PowerUp essentially does all this for your. It uses a model in which the powershell (well psake) scripts can move execution to another machine by calling a specific function. This can either be done with powershell remoting or psexec - you wouldn't need to change the script, it just requires a setting per environment to say which you would like to use.
Check other the sample at https://github.com/AffinityID/PowerUpSamples/tree/master/SimpleWebsite.
Hopefully that shows you enough, but if not let me know and we can go into more detail.

Running a cgi perl script as an Administrator

I'm writing a perl script for a website, and I need to be able to control VirtualBox via the website. I'm not sure where to start, or if I'm even trying to debug in the right area, but here goes.
My server is running IIS7 on Windows Server 2008 R2. I'm also running 2 virtual machines through the vboxmanage command line interface. These VMs are running under SERVER\administrator.
When I open my website, it requests a login. I login to the website as SERVER\administrator and click a link that calls my script using an xmlhttprequest. Now, normally, it doesn't matter what user I run these as, but with vboxmanage, if I run the command as a different user, the list of VMs is different. I tried whoami, which returned SERVER\administrator, but %DOMAINNAME%\%USERNAME% returns the domain that the server is connected to as dommainname and SERVER$ as the username. The vboxmanage command then fails.
On the website, impersonation is turned on. When I turn impersonation off, the whoami request changes to be iis apppool\website. Any ideas on how to get around this?
As a final note, I've thought about using runas, but since it prompts for a password, there's no way to call it through scripting (and that would be a poor security decision, I'd imagine).
This is an oft recurring, well-known and well-solved problem. Instead of having one big program dealing with requests from the Web and managing the VM (strong coupling), separate the concern and write two programs, each doing exactly one task.
The user facing program running in the Web server context can continue with limited privileges. The VM manager is a stand-alone program running with the necessary admin privileges, either repeatedly from the scheduler or as daemon/service.
Have the first communicate with the second over a message-queue.