In general I'm loving ZeroRPC, working great and has been for months (see https://github.com/SuperCowPowers/workbench for a kewl project on top of ZeroRPC). But there's a small gap in my understanding that I wanted to ask about.
I'm making the following two assumptions (which could be wrong):
- If you have any problem with heartbeats you have a server side gevent/greenlet issue and you should gevent monkey patch it (i.e. from gevent import monkey; monkey.patch_all()). BTW this is working great for me. :)
- If you have any problem with timeouts (heartbeats are working fine), then your server task is running longer than 30 seconds and the client is properly saying 'Hey this is taking too long I'm going to pull out', which is good.
So if those two assumptions are correct, I have a 'usage' question. My server will have a reasonable number of tasks that will run longer than 30 seconds, because I'm encouraging people to write their own clients (it's SO easy) should I just dogmatically document that folks should set the client timeout option to something higher?
My only hesitation about that path is that obviously people won't do it, it will run fine on small stuff and then timeout on bigger stuff.
