"pm2 reload" while reload in progress leads to zombie processes and pm2 errors

[migg24]

What's going wrong?
When you send pm2 a pm2 reload app while a pm2 reload app is still in progress it leads to errors and zombie processes of the app that are not visible in pm2 list but in ps aux | grep node. The pm2 daemon IMHO should not reload the app if a reload is in progress.

How could we reproduce this issue?

1. Start an app with 8 instances, give it a kill timeout of 10000 and a 5000 timeout on SIGINT to exit the process.
2. Stream pm2 logs
3. Open 2 more terminals on the same machine
4. Execute pm2 reload app on terminal 2, and shortly after that execute pm2 reload app on terminal 3

In my test it resulted in this error:

[PM2][ERROR] Process 6 not found

And the pm2 reload stopped, but this error starts to fill the pm2.log:

Unknown id _old_6

After that there are 8 instances in pm2 list but more than 8 in ps aux | grep node and they seem to still respond to connections so that some are working on an old version of the code and some on the new version if there is a code change that triggered the reload.

Supporting information

PM2 version: 2.5.0, 2.1.6
Node version: 4.8.3
Linux

I think the only option to handle this is to no let the pm2 god daemon reload if it is already reloading. But maybe there are other ideas/options.

[Unitech]

pm2 reload lock via timestamped lockfile added, to try it:

$ npm install Unitech/pm2#development -g
$ pm2 update
# Try to start two pm2 reload in same time  
$ pm2 reload all & pm2 reload all

[kschingiz]

@Unitech It didn't helped me, it created another instance of app, so in pm2 status it shows 2 instances, but node processes 4, was 3 before you instructions.
I have 2 instances with latest update of app, but 2 is deprecated, and my clients redirecting to older instances some time, it creates me bugs and crashes of mobile apps.
How can I solve it?

[Unitech]

Published on PM2 2.6.0:

$ npm install pm2@latest -g
$ pm2 update

[migg24]

Big thanks for this quick fix! We tested it in our environment and the problem does not appear anymore.
👍