It depends. You are going to have to just learn about Channels
Ignore Stdio and friends, they are distracting you. What you have is a Channel... how you got that channel is irrelevant... I have an implementation of Channel (in closed source product) that doesn't even have a persistent connection between the two endpoints... though most "normal" channels are established using either the "JNLP" protocol family or the SSH mechanism. At its fundamental, Channel sends Callable instances from one side to the other where they get invoked. The sending involves serialisation, so it is all pass-by-value. Because that would be somewhat limiting, you can mark some objects as "exported" (providing that they implement an interface to proxy) when you do that, remoting replaces the object with a proxy, where each method call will result in a request being sent over and back... so exported objects are pass-by-reference There are a few other special case objects: I/O streams mostly, and also classloaders For the I/O streams, a more intelligent "proxy" is used (as you cannot proxy classes, only interfaces) which does things like chunking of data to improve performance For the classloaders, there is some deep voodoo in order to allow reconciling which classloaders to resolve classes from in order to make everything seem seamless. HTH On Thu 29 Dec 2016 at 00:55, Doug DesCombaz <[email protected]> wrote: > I'm trying to find some documentation of how Jenkins shares information > between master and slaves. I'm not necessarily interested in the channels, > or mechanisms, but the guarantees (wishes) on how the memory would be > distributed (I am also interested in the channels, and mechanisms, but > that's another question). > > So, I know (or rather I think I know) there are two connections, a > bi-directional tcp, and a bi-directional of standard io passed through the > ssh connection. This is what I keep finding when asking the question, but > it's not quite the answer that I am interested in. So, I will ask w/ an > example: > > In master I have a map of data ([ somekey: "some value"]). I make this > data available to a closure that I am passing to slave executors slave-a, > and slave-b. If slave-b modifies this map (or it's local copy of it), 1) is > the change reflected in the master?, 2) is the change reflected on the > other slave-a?, 3) If yes, what are the guarantees (or (un)documented > behaviors) here? > > I don't know, but I suspect that my self-defined objects are not > synchronized throughout the "cluster" (especially since Jenkins distances > itself from being considered a cluster). Is that true? I wonder however if > properties defined in "env" are (simply because it's a data structure under > their control)? > > Anyway, if you are able to answer any of those, maybe you can describe how > it's being done too just to satisfy my curiosity. > > > > > > > > > > > > > -- > > > You received this message because you are subscribed to the Google Groups > "Jenkins Users" group. > > > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > > > To view this discussion on the web visit > https://groups.google.com/d/msgid/jenkinsci-users/2f600c3a-fd9f-4b50-a026-53a3faacc288%40googlegroups.com > <https://groups.google.com/d/msgid/jenkinsci-users/2f600c3a-fd9f-4b50-a026-53a3faacc288%40googlegroups.com?utm_medium=email&utm_source=footer> > . > > > For more options, visit https://groups.google.com/d/optout. > > > -- Sent from my phone -- You received this message because you are subscribed to the Google Groups "Jenkins Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/jenkinsci-users/CA%2BnPnMwYxhFQOCkTsGwQj--eHC9fkV%2BBcyg6vDHU7fgV8_H%3DNg%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
