[HTCondor-devel] DaemonCore ChildAlive & systemd/docker/etc?

[Erik Paulson <epaulson@xxxxxxxxxxx>]

[I've been ignoring a lot of the Linux universe for a couple of years, so this message might be a bit non-sensical]

So, one of the nice things about DaemonCore is the regular heartbeat message that the daemons send to the master - the affirmative heartbeat is better than simply asking the kernel "is this process still there?" - if the master gets a message and can properly decode the message, it's a good sign that the child daemon is at least still semi-functional.Â

It's a shame that this hasn't been standard in operating systems since V7 Unix or something.Â

It looks like Systemd kinda does this today, with sd_notify:Â

https://plus.google.com/+LennartPoetteringTheOneAndOnly/posts/37AWJLE3XcJ

Does it make any sense for Condor to set the systemd environment variables and implement the systemd notification protocol, so if the master wanted to manage a non-condor daemon it would just work?

Do any of the other management-y tools like docker use something similar? (I ask for a couple of reasons, one of which is that it sounds like some of the docker management system is coming to Windows, so if there was some sort of heartbeat protocol that docker offered processes it might be a mechanism that Condor could reuse across Linux and Windows, too)

-Erik