Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!CC3.CC.UMR.EDU!obrennan From: obrennan@CC3.CC.UMR.EDU (obrennan) Newsgroups: comp.sys.apollo Subject: RE: Registry problems Message-ID: <9012261430.AA03164@cc2.cc.umr.edu> Date: 26 Dec 90 14:30:13 GMT Sender: daemon@ucbvax.BERKELEY.EDU Organization: The Internet Lines: 47 llbd, glbd and rgyd seemed to be running fine, but then I tried edrgy, and it didn't start (unfortunately didn't save the error message). Thougth, aha, maybe rgyd is trashed and killed it... Shouldn't have done that, because when trying to restart it I get: > Registry: Fatal Error - Cannot register server replica interface - 0X1C010001 - > communications failure (network computing system/RPC runtime) > Registry: Fatal Error - 0X9010006 - IOT instruction fault (UNIX/signal Sounds like the glbd is not 1) getting started before rgyd, or 2) accepting server requests to register. We have had troubles in the past when placing the GLBD and RGYD on the same node. Apparentally, GLBD is a real memory hog and the node spends much of its time paging. I would try rebuilding the GLBD with -create -first options and see if that clears things up. Also, if you decide to move GLBD to another node, make sure that node is up before the RGYD node so that the RGYD can register with the GLBD (need LLBD on both nodes). If you have multiple GLBDs there are other considerations when rebuilding the GLBDs (see Managing NCS software). Arghh! I'm no apollo expert, so I just rebooted the machine again. This time "rgyd" didn't start even on boot. It didn't complain when I started it manually, but it seems non-functional (hmm.. just now when I tried, it seems to have gotten something ok, could "su" again normally, but still cannot run "edrgy", I get... When you manually start the RGYD are you using /etc/server -p to start the RGYD as root; otherwise the registry server won't run with the needed rights. ># edrgy >?(edrgy) Unable to open the registry. - Registry server unavailable (RGYC/Serve >r) ># This is due to the initial problem. Hope this helps. Gerry O'Brennan Programmer/Analyst II Computing Services University of Missouri - Rolla ------------------------------ obrennan@apollo.cc.umr.edu c0022@umrvmb.umr.edu ------------------------------