I have an akka-scala app that I'm able to build as a jar. I can then send the jar to a remote server via Jenkins. However, I don't know how to properly deploy the app.
The first problem is killing previous instances of my app. If in Jenkins I try to execute such commands as pkill -f %proc% or ps ax | grep ... | awk ... | xargs kill -9 via ssh, Jenkins finishes with return code -1 even if I add || true to the end of these commands.
The second problem is starting the app. I'm able to successfully run the jar with nohup java -jar ... & command, but it doesn't really start though I'm able to see it in the process list.
None of these problems occur if I execute commands manually (even via ssh).
My question is - what am I doing wrong? What is the proper way to perform such task? Am I using the wrong tool?
Consider using something like supervisord for starting/stoping your app and also restarting it in case of crash or server reboot.
Related
Using Postgres 9.5 and the libpqxx c++ bindings, I want to launch a copy of postgres that is not installed on the users machine, but is instead packaged up in my application directory.
Currently, I am using pg_ctl.exe to start and stop the server, however when we do this, pg_ctl.exe seems to launch postgres.exe in a new terminal window.
I want it to launch postgres.exe in a headless state, but can't work out how.
I have tried enabling/disabling the logging collector, setting the logging method to a csv file (instead of stdout/stderr), and a couple of other logging related things, but I don't think the issue is the logging.
I have also tried running postgres.exe manually (without pg_ctl) and can get that to run headless by spawning it as a background process and redirecting the logs, but I would prefer to use the "pg_ctl start" api for the "wait for startup" (-w), and "timeout" (-t) options that it provides.
I believe you won't be able to do that with pg_ctl.
It is perfectly fine to start PostgreSQL directly through the server executable postgres.exe. Alternatively, you can use pg_ctl register to create a service and start the service.
In my use case, I was able to resolve the issue by running pg_ctl.exe using
CreateProcess, and providing the dwCreationFlags CREATE_NEW_PROCESS_GROUP | CREATE_NO_WINDOW.
I was originally using CREATE_NEW_PROCESS_GROUP | DETACHED_PROCESS, but DETACHED_PROCESS still allowed a postgres terminal to appear. This is because DETACHED_PROCESS will spawn the pg_ctl without a console, but any process that inherits stdin/stdout from pg_ctl will try to use it's console, and since there isn't one, one will be spawned. CREATE_NO_WINDOW however will launch the process with a conhost.exe, however the console will have no window. When the executables spawned by pg_ctl try to write to the terminal, they will successfully write to the console created by the conhost.exe which has no window.
I am now able to run pg_ctl from code with no console appearing.
I have an ansible playbook which deploys a jboss eap instance alongside some other things. The playbook runs fine until it gets to the point to start jboss using the provided standalone.sh script. I am using the shell module to start this script and it works fine in that jboss starts however when the task is executed ansible does not return any status message like a changed or OK and just seems to hang.
Is there a way I can force ansible to see this as something which has changed the system state ?
I don't personally use jboss, but this sounds to me like the startup.sh script simply isn't launching jboss in the background, so ansible is simply waiting for it to end and it never does.
There are a few potential ways you can address this. Given the information in this thread you might want to try a task like this:
- name: start jboss
shell: nohup standalone.sh > /dev/null
async: True
poll: 0
The poll: 0 tells Ansible to 'fire and forget' the command. If you don't include this then Ansible will kill the process when it returns from the remote server.
Another possibility is to use an init script. The thread I linked to above points to a location where you should be able to find an init script. You could install that (and leave it disabled if you don't want jboss to start up when the system reboots), and then simply run it via the service command from Ansible:
- name: start jboss
service: name=jboss state=started
Edit: If you're unwilling or unable to use nohup or an init script that puts jboss into the background then another alternative is to make use of screen if you have that installed and available to you. I regularly do this when invoking long-running commands from Ansible so that I can check back well after the playbook has been run to check on the status of the command:
- name: Invoke long running command
command: /usr/bin/screen -d -m /path/to/my/command
async: True
poll: 0
This will launch a detached screen session and invoke your command inside it. Ansible will promptly forget about it, leaving the command to run on its own.
I am using Mirth connect which uses install4j to launch the program.
I am using the mcservice program and would like to get the pid of the launched application so that I can monitor it. How do I do this? Right now the service only has the standard start, status, etc commands.
Back in 2011 there seemed to be some indication that pid monitoring would be coming soon: http://www.mirthcorp.com/community/forums/showthread.php?t=5509
If you need get the the number of the process pid, you can do it by the terminal command:
$ pgrep mcservice
Also, you can store the pid number into a file and run the next command:
$ kill -9 `cat /path/to/file.pid`
In this case, you choosing the best solution for your issue. If you need something more complex you can view this link: Simple Process Checker To Find Out If A Service Is Running or Not
here's my situation.
I'm running Cruise Control as a Windows Service and trying to get it to connect to a Mercurial Repository on BitBucket over SSH.
I'm pretty sure that everything's configured OK (PuttyGen, Pagaent, etc). I'm remoting onto the server using the same account that I am using to run the service and if I issue hg pull -b ssh://#bitbucket.org// from a command line everything works. I added -v to the ssh configuration in mercurial.ini and I can see all of the steps that are taken.
If I run CC.NET from a command prompt then it builds fine. In the console window I can see the same logging from the SSH operation.
However, if I run CC.NET as a service (using the same user account that I'm logged in on) the call to BitBucket times out. I can find no way to work out why either. The build log doesn't help and neither do ccnet.log or ccnet.trace in the temp directory. I was expecting one of them to contain the logging from the SSH operation, but they don't.
Can anyone help? Is it that running as a service prevents it from connecting to Pagaent (I've started Pagaent by adding it to HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Run). When I did the pull from the command line I had to OK a dialog, but only once. Is it waiting on the same dialog now that it's running as a service?
Getting close to my wits end here.
Thanks
I did get it working in the end. The trick was to create the public key without a passphrase. When running as a service the solution has to be completely non-interactive and the passphrase option with pagaent.exe just isn't.
Here are the steps:
Use PutTTygen to generate a secure key WITHOUT a passphrase. If you really do need one then you can add it to the mercurial.ini file, but defeats the point for me as it's in plain sight anyway.
Copy a mercurial.ini to two locations: C:\Windows\System32\config\systemprofile and C:\Windows\SysWOW64\config\systemprofile. Probably only one of these was really necessary, but I didn't have the time to experiment. The first is the home directory for the system user when running 64 bit apps, the SysWOW64 location for 32 bit. Make sure that if you do the same as me then keep both files in sync - or go one further and work out which is the correct location.
Add something like this line under the [ui] key in both files:
ssh = "D:\Program Files\TortoiseHg\TortoisePlink.exe" -ssh -2 -C -batch -v -i "[Path to your ppk file]"
Add the passphrase to the end of the command if one was created in step 1.
Make sure that TortoisePlink.exe is specified, not Plink.exe. They should both be in the same directory.
Download psexec from http://technet.microsoft.com/en-gb/sysinternals/bb842062.aspx
Run d:\PSTools\PsExec.exe -s -i cmd.exe. This will open a command line as the system account in interactive mode.
Now do an hg pull, or hg clone or whatever.
A dialog should pop up with a confirmation message. This is a one time thing and the reason that you have to do the PsExec step. OK the dialog.
Now cc.net should be able to be run as a service under the local system account using SSH!
I'm trying to deploy a play2 application with capistrano, but I can't figure out how to (re)start the play2 application after a successful deployment. Just triggering 'play start' will cause the process to be hanging waiting for me to press ctrl+D
I've created a start script in the play app root folder
#!/bin/bash
nohup bash -c "/var/lib/play2/play start &>> /tmp/myapp.log 2>&1" &> /dev/null &
It works great when I run this on the server. When I try to call this from my local machine over ssh it also works. But when I am using capistrano, it doesn't seem to do anything. My capistrano config looks like this:
namespace :deploy do
task :restart do
stop
sleep 1
start
end
task :start do
run "cd #{current_release}/trip-api && ./start.sh"
end
task :stop do
run "cd #{current_release}/trip-api && ./stop.sh"
end
end
What's the best way to start a play2 application on a remote machine? How to get it working with capistrano?
Have a look at play documentation on deploying your application on production
The recommended way is to package your app with
play clean compile stage
And then run it with
$ target/start
To stop it, have a look at the docs:
The server’s process id is displayed at bootstrap and written to the
RUNNING_PID file. To kill a running Play server, it is enough to send
a SIGTERM to the process to properly shutdown the application.
In this quickstart for Openshift, it shows another way to start play as a service and how to stop it.
basically you do something like this to start:
APP_COMMAND="${OPENSHIFT_REPO_DIR}target/start $PLAY_PARAMS "\
"-Dhttp.port=${OPENSHIFT_INTERNAL_PORT} "\
"-Dhttp.address=${OPENSHIFT_INTERNAL_IP} "\
"-Dconfig.resource=openshift.conf"
echo $APP_COMMAND &>> $LOG_FILE
nohup bash -c "${APP_COMMAND} &>> ${LOG_FILE} 2>&1" &> /dev/null &
and to stop it
pid=`cat RUNNING_PID`
echo "Stopping play application" >> $LOG_FILE
kill -SIGTERM $pid
There are few fresh topics about running application available at Google Groups:
Start an application as a background process
When deployed on Ubuntu 10.04 cant detach from console
It's good idea to follow or join them
I would suggest using runit. We are currently running a bunch of services in production and it works great.
It only involves creating a simple shell script named run, pointing runit to its containing directory and then start it. Services should not daemonize by themselves and runit controls pid files, etc.
There is a command ( sv ) to start, stop and query services. ( sv start|stop|status|restart yourapp ).
A cursory google search got me this http://rubygems.org/gems/capistrano-runit though I do not use capistrano at all so I can't vouch for it's usefulness.
http://smarden.org/runit/
The faq is the best place to start: http://smarden.org/runit/faq.html
In debian you just apt-get install runit and are good to go.
update-service --add /your/service/dir/ will register the service with runit.
On deployment we stop services, change binaries and start services; it is really simple.