Agent respawn upon death

[MikeHawkesCapventis]

Proposed change

If the agent dies for any reason, there's no way to respawn it - the nex node dies with it. The only recovery is to restart nex, and then redeploy (even though the scripts / binaries are already in the NEXCLIFILES object store bucket). If the agent some way of recovering from a critical failure, it would make nex a lot more resilient.

Use case

Most things would work nicely, however, I could write a script that causes the runtime engine to crash out (perhaps by consuming too much memory, or a network timeout issue when connecting remotely). If nex could trigger an agent restart and redeploy the things that were on the failing node, it would help.

Contribution

No response

[jordan-rash]

In the new version of Nex, agents are treated the same as services. They have a restart count associated with them, I believe 3 by default. If your agent fails more than 3 times, then probably something that needs to be addressed in the agent code...

[MikeHawkesCapventis]

Whoohoo! It's the one thing that's caught me out a few times - I added Goja and Anko to test out. While I wrapped with a recover, I couldn't stop the agent from dying if I did something stupid. This sounds like it's fixed - so should I close this?

[jordan-rash]

Let's leave it open until you can play with it and let me know what you think. Admittedly, the V3 will probably merge today, but the agent sdk might be another week or so, so you might not be able to verify its behavior is what you are looking for a few more weeks. I am keen to hear your feedback, so lets leave this open