Checkpoint and Restore

[juancampa]

Hi, I was reading through @hashseed's proposal in #17058 and started wondering if this could be used to "checkpoint & restore" a running node program. I understand that a process is more than just the isolate and contexts (i.e there are OS-level resources like file descriptors, sockets, etc that are not serializable). But assuming that it would be acceptable to lose that part of the state. Is this something that is currently possible? would this be enabled or facilitated by the #17058 proposal?

I was picturing something like: send a signal to a running node.js process (i.e. kill –USR2 3049) and cause it to: wait until the current callstack is finished, serialize the state into a file, and exit. Later the same program could be restored by providing the file to a new process. Of course any file/network IO would have to be manually restored at this point.

Basically something like CRIU but specifically for node.js which could make it potentially faster and space efficient.

[refack]

Hello @juancampa,

Check out the discussion in #13877

[juancampa]

Thanks @refack. I also asked on #17058 and got a direct answer here: #17058 (comment)