dj@dorsai.cognet.ucla.edu (David J. Wells) (11/19/88)
Environment: DN4000 (dorsai.cognet.ucla.edu) SR10.0 Domain/OS BSD 4.3 running tcpd, syslogd, inetd, llbd, glbd, rgyd, cron, lpd, netman, sendmail. shell: /bin/ksh Description: If I have 13 su's (root), then applications cannot communicate with the rgyd. "Why do I want 13 su's?" Well, that was just a way that I could duplicate the problem under DM -- I first saw the problem under X11R3 Windows. Under X, I am allowed 3, maybe 4 su's before the same thing happens. Apparently some X programs use the same resource that is required to communicate with the rgyd. (xterm is *not* one of these affected programs) When this happens, many things break. Most noticeable are su, login, rgy_admin, and sendmail (it doesn't know about any local users). ~ $ su Su failed ~ $ rgy_admin Cannot locate a rgy replica which is in service Default object: rgy default host: unspecified rgy_admin: q ~ $ rsh dorsai (from dorsai) account is invalid or has expired (RGYC/Login) Login incorrect Connection closed. ~ $ Eventually, after I kill an su, everything starts working again. Oddly, there can be a time where I can su in one window, but not in another; yet I can run rgy_admin in any window. If I wait 5 minutes or so, then it all works. ~ $ rgy_admin Default object: rgy default host: dds://fs3 State: in service master replica list is readonly rgy_admin: lr (master) dds://fs3 dds://dorsai rgy_admin: q ~ $ Attempted Solutions: It was suggested that I might be running out of processes. Since I can start more processes, this is not the case. I thought that maybe I need more server processes, so I tried netsvc -s 3, but this didn't help. My next idea was that I might be running out of a socket-related resource. But I can start rlogin sessions to other machines, so that isn't the problem. Any other ideas? Thanks in advance. David J Wells dj@cs.ucla.edu w213/206-3960
pato@apollo.COM (Joe Pato) (12/06/88)
In article <18085@shemp.CS.UCLA.EDU> dj@cs.ucla.edu (David J. Wells) writes: > >Environment: > > DN4000 (dorsai.cognet.ucla.edu) > SR10.0 Domain/OS BSD 4.3 > running tcpd, syslogd, inetd, llbd, glbd, rgyd, cron, lpd, > netman, sendmail. > shell: /bin/ksh > >Description: > > If I have 13 su's (root), then applications cannot communicate > with the rgyd. "Why do I want 13 su's?" Well, that was just a > way that I could duplicate the problem under DM -- I first saw > the problem under X11R3 Windows. Under X, I am allowed 3, maybe > 4 su's before the same thing happens. Apparently some X programs > use the same resource that is required to communicate with the > rgyd. (xterm is *not* one of these affected programs) > > When this happens, many things break. Most noticeable are su, > login, rgy_admin, and sendmail (it doesn't know about any local > users). > . . . > >Attempted Solutions: > > It was suggested that I might be running out of processes. Since > I can start more processes, this is not the case. > > I thought that maybe I need more server processes, so I tried > netsvc -s 3, but this didn't help. > > My next idea was that I might be running out of a socket-related > resource. But I can start rlogin sessions to other machines, so > that isn't the problem. > > Any other ideas? > > Thanks in advance. > David J Wells > dj@cs.ucla.edu > w213/206-3960 There is, in fact, a socket related resource that is not properly released in sr10.0. IP family sockets are not affected, which is why you were able to continue to create new rlogin sessions. This problem has been fixed in sr10.1 Joe Pato UUCP: ...{attunix,uw-beaver,brunix}!apollo!pato Apollo Computer Inc. NSFNET: pato@apollo.com