I concur that running two independent agent directories is preferable, because of the ability for independent updates as you pointed out (and also, not absolutely neccessitated but often going along, the prevention of log file overwriting/confusion fubars because of, then, separated tmp/ and out/ directories). OP was initially asking for the most lightweight setup, but splitting agents has undeniable benefits over the most slim setup.I'm not convinced on the value of two smgr yet. Yes, it can easily be done, but I guess you'd need two tcp ports, with firewall implications etc, multiple init scripts, more monitoring etc., and not 100% sure if smgr-dialog can connect to anything but tcp/8871?We here treat smgr as the single instance starter for all Automic processes on one machine. What we do here (historically, not my original idea) is use the Phrase UC4P for prod, UC4T for test and UC4X (or just UC4) for "multi environment systems".But yeah, I've seen varying setups that work, and that's also the result of Automic always leaving customers without a proper reference setup imho. But from the design of the smgr, I always gravitate towards one smgr, multiple agents myself.n.b. Ultimately, I always considered smgr to be the eqvialent to Daniel J. Bernsteins daemon-tools: born mostly from "not invented here" syndrome with the init system :)Br,
I sometimes point out that keeping them separated makes my OCD happy
Brian said: "...SMGRGui, as once they have access to the DEV, the user can simply change over to the PROD and bring it down."