Custom App not starting automatically after installation/reboot [MTCAP-L4E1]

[Rafael]

Hi,

We have a node application that runs on multiple MTCAP-LEU1 (firmware v1.7.4) as a custom app. But now we’re having problems when installing it on a MTCAP-L4E1 (firmware v5.0.1-AEP) using deviceHQ.

I can see that the app files were downloaded in /var/config/app/app-name but the application is not running automatically after restarting the gateway. It works as expected on MTCAP-LEU1 (v1.7.4) and we didn’t change the code.

I can run the app manually with the following commands:

./Start start

sudo ./Start start

sudo app-manager --command start --appid <APPID>

Output of sudo app-manager --command status:

{
	"Apps" : 
	[
		{
			"AppDescription" : "...",
			"AppId" : "<APPID>",
			"AppInfo" : "Installed by -",
			"AppName" : "app-name",
			"AppType" : "CUSTOM",
			"AppVersion" : "8.3",
			"AppVersionNotes" : "...",
			"Enabled" : true,
			"status" : null
		}
	]
}

Output of sudo app-manager --command list-installed:

{
   "Apps" : []
}

Logs in /var/logs/messages:

annexcd: [INFO] delivery_agent.cc:RunnableCore:3437: sending active apps                                                              
annexcd: [INFO] mts_device_interface.cc:getActiveApps:541: Node-red disabled so skipping node-red apps                                
after-switch-inet: restarting annex client                                                                                            
APPMANAGER: AppCommand::executeStatus: Getting app statuses                                                                           
APPMANAGER: AppCommand::getBasePathFor: unknown app type. Defaulting to custom                                                        
APPMANAGER: AppCommand::executeStatus: can't read pid as a string, trying as number:#012 in Json::Value::asCString(): requires stringV
APPMANAGER: Exiting with success status                                                                                               
annexcd: [ERR] delivery_agent.cc:FillActiveApps:1042: field [status] missing from ActiveApp  - skipping                               
annexcd: [INFO] delivery_agent.cc:SendActiveApps:987: Sending Active Apps data

Permissions of the custom app files:

-rwxrwxrwx 1 root root

Permissions of the config file:

-rw-r--r-- 1 root root

And I can see our application in the list of custom apps in the gateway UI (custom apps are enabled).

Any ideas on how to solve this problem?

Thanks,

Rafael

[Jeff Hatch]

Hello Rafael,

Could you do a “cat /var/config/AppManager.json” and let me know what is in that file?

Thank You,

Jeff

[Rafael]

Hi Jeff,

Here it is:

admin@mtcap:~$ cat /var/config/AppManager.json
{
	"Apps" : 
	[
		{
			"AppDescription" : "...",
			"AppId" : "<APPID>",
			"AppInfo" : "Not Available",
			"AppName" : "app-name",
			"AppType" : "CUSTOM",
			"AppVersion" : "8.3",
			"AppVersionNotes" : "...",
			"Enabled" : true
		}
	]
}

Thank you,

Rafael

[Jeff Hatch]

Hello Rafael,

It doesn’t appear that the SDCard value is defined in the application manifest file, so it should be installed on an SD Card. Is that correct? We need to verify where it got installed to.

Do you have any logging from when you ran the “sudo app-manager –command start –appid ” command? The only output I see that you captured from app-manager looks like it is running the status command and having difficulty finding the status.json file. Possibly because the app isn’t where it expects it to be.

Jeff

[Rafael]

Hi Jeff,

We don’t use an SDCard and the application is installed in: /var/config/app/app-name.

Output of the app-manager start command:

admin@mtcap:~$ sudo app-manager --command start --appid APPID
Password: 
AppCommand::executeStartScript: executing: /var/config/app/app-name/Start start

After that, the application runs correctly and there are no more logs. But it doesn’t run in the background. After running the start command, I can’t use the terminal anymore. I don’t know if that’s normal, in the Start script we run the node app using nohup.

Thanks,

Rafael

• This reply was modified 4 years ago by Rafael.
• This reply was modified 4 years ago by Rafael.
• This reply was modified 4 years ago by Rafael.

[Jeff Hatch]

Hello Rafael,

In the Start script, what does the command that starts the application look like?

Jeff

[Rafael]

It only runs another shell script:

function ExecuteStart {
    echo "ExecuteStart:"
    ./run-app
}

And run-app looks like this:

nohup node app.js &

Thanks,

Rafael

[Jeff Hatch]

Hello Rafael,

I think that by executing the “run-app” script in the foreground it may be what is causing the app to not go into the background when executed at the command line. Try running “run-app” in the background, ie “run-app &”. Let’s see if that “fixes” the execute at the CLI behavior.

It is odd, but I have seen some interesting behavior between processes and deadlocks with non-backgrounded processes and also with running backgrounded process where the app’s pipes (stdout and stderr) were not redirected to /dev/null. By trying to run with the something like the following: “node app.js > /dev/null 2>&1 &” it might help cause all the pipes to be “disconnected” between the parent and child processes.

Jeff

[Rafael]

Ok, so I did what you said: run-app & in the Start script and node app.js > /dev/null 2>&1 & in the run-app script. And now, after starting the app with app-manager, it runs in the background:

admin@mtcap:/var/config/app/app-name$ sudo app-manager --command start --appid APPID
AppCommand::executeStartScript: executing: /var/config/app/app-name/Start start
Success!
admin@mtcap:/var/config/app/app-name$

And these are the logs in /var/log/messages:

2020-10-30T19:24:26.319473+00:00 mtcap APPMANAGER: AppCommand::getBasePathFor: unknown app type. Defaulting to custom
2020-10-30T19:24:26.319976+00:00 mtcap APPMANAGER: AppCommand::getBasePathFor: unknown app type. Defaulting to custom
2020-10-30T19:24:26.320385+00:00 mtcap APPMANAGER: AppCommand::getBasePathFor: unknown app type. Defaulting to custom
2020-10-30T19:24:26.333177+00:00 mtcap APPMANAGER: AppCommand::executeStartScript: executing: /var/config/app/app-name/Start start
2020-10-30T19:24:26.477978+00:00 mtcap APPMANAGER: AppDataDB::SetAppData: updating app info in /var/config/AppManager.json with appId: APPID
2020-10-30T19:24:26.486483+00:00 mtcap APPMANAGER: Exiting with success status

But the application still doesn’t start after rebooting the gateway (our main problem). Any other ideas on how to solve this problem?

Thanks,

Rafael

[Jeff Hatch]

Hello Rafael,

I did find one interesting difference between starting the app by hand at the command line and through the init script on boot. On boot the /etc/init.d/customapp script sets the –initd flag when executing the app-manager program. This would have the effect of causing the Start script to get executed as “/var/config/app/app-name/Start start –initd”. If that messes up the argument processing on the Start script maybe that is causing a problem.

Jeff

[demetre@swudutchyy.com]

Looking for more information and we get from this site. https://slimsoupdiet.com/

[Author]

Posts