Ping does not succeed and causes action to time out frequently

[slaughtr]

I've just implemented this action in several repos, and for the most part it has been a great thing compared to the self hosted runner spiderweb. However, something like 10% (maybe even 20%) of the time, the ping never succeeds and the action fails. My devs are getting really annoyed about this.

I've set the ping to two different hosts (the most static ones in my tailnet, subnet routers) that are deployed in AWS EC2 in multiple accounts. These hosts work fine on their own, I'd be very aware of issues with them. I don't think they're the problem but maybe they are?

When the ping succeeds, its usually in ~5 seconds. Sometimes just one ping and ready to go.

I'm considering deploying some other static host(s) just to offset this, but then I'm managing infra just for this github action which is what I was more or less trying to get away from.

Definitely feels like I'm doing something wrong. I'd love to be able to tell my devs they don't need to hit Re-run failed jobs anymore.

Any suggestions?

[slaughtr]

Just to really underscore how frustrating this can be, I've been trying to deploy a change to production for 30 minutes. I've re-run this like 8 times. Same two hosts that worked fine for all other deploys today, but for this one just can't get it to go. I feel like this can't be how it should work.

[slaughtr]

Well after some experimenting and double checking the code, I now understand that the ping must hit all hosts successfully first. And that subnet routers aren't good targets, for whatever reason. And that you don't have to use a machine in your tailnet, just something in your tailnet that can respond to ping.

I feel silly but I do feel like the documentation is rather fuzzy here.

EDIT: nope this still fails very frequently. About as often, maybe slightly less. Makes sense it wouldn't change much.