Fork me on GitHub

Can Onyx lifecycles be parameterized based on runtime dependencies (possibly injected) into the peers' component / system?


Like a database connection determined when the system map is built, which needs to be used by a custom output plugin.


@aengelberg: I'm sure there are ways to achieve what you want, since they can execute arbitrary code. It might be a little awkward, but it should be doable


subscribe-to-log calls 2 methods - find-job-scheduler and find-messaging This methods use

(do (Thread/sleep 500)
when there is problem with ZK connection. So you will have long log file (produced by (catch Throwable e (warn e)) and hanging point when there is connection problem for a long time. I think that it is better to have fail fast methods rather than hanging points all over the place. This is also annoing when I am in the repl and I turn off ZK - because of logged exception for each 500ms.


I think they should fail fast if the connection has been lost, but they should retry if the node doesn’t exist


How suitable is Onyx for CQRS type applications, where domain objects handle certain command messages and emit event messages for event sourcing?


@yonatanel very suitable. Someone wrote a blog post about it way back


I got an email last week about a student possibly picking Onyx's masterless architecture as a topic for his/her master's thesis. Would be pretty cool to see ideas disseminate back into academia from industry since a lot of Onyx's design comes from research papers to begin with.


Definitely. It’d be good to see what they can come up with too


While we're on the topic, if anyone is/knows of a student that's interested in studying or publishing on this stuff, we're happy to make the time to collaborate. cc @dignati 🙂