These are chat archives for fanout/pushpin

10th
Apr 2018
Randy Watler
@rwatler
Apr 10 2018 15:43
Hello... new Pushpin user here. I am wondering how to incorporate multiple Pushpin instances locally for fail over and limited scale short of something like Fanout cloud. It seems that chaining is a preferred method to scale out, but I am wondering why the GRIP protocols in that case are routed through the "closest" Pushpin instance? Is it possible to simply have more than one load balanced Pushpin instances running and broadcast all publish messages to each? If so, I assume that this is the responsibility of the back end server as documented in the proxy chaining section, correct?
Justin Karneges
@jkarneges
Apr 10 2018 16:54
hi @rwatler , it is indeed possible to have more than one Pushpin instance behind a load balancer, and publish data to all of them. this is exactly how it is meant to be scaled
chaining is sort of a separate thing. that's for delegating between different pushpin tiers, presumably because you want to leverage greater scale of a more further out tier. but the way that further out tier itself would scale would be by having multiple Pushpin instances behind a load balancer :)
Randy Watler
@rwatler
Apr 10 2018 17:49
@jkarneges Great. We are building a service around Pushpin and I wanted to make sure I had the general contract down. I suppose then that a "smart" layer of such could be constructed to publish only to the specific Pushpin instances that had relevant subscriptions, but for small N I'm guessing that it would be fine to simply broadcast the publishes to all.
Justin Karneges
@jkarneges
Apr 10 2018 19:06
@rwatler yup, unless/until you're sending so much data that the duplicate sends are impacting performance, I suggest just sending to all. anything smarter than that will make debugging harder
but when it comes time to do that kind of thing, pushpin offers ways
Randy Watler
@rwatler
Apr 10 2018 19:34
@jkarneges Thanks!