Path: utzoo!attcan!uunet!tut.cis.ohio-state.edu!ucbvax!CORNELLC.CIT.CORNELL.EDU!pschenk%cernapo.cern.ch From: pschenk%cernapo.cern.ch@CORNELLC.CIT.CORNELL.EDU (P.R.Schenk/EP) Newsgroups: comp.sys.apollo Subject: Hung TCP/IP revisited Message-ID: <9006281544.AA05152@dxmint.cern.ch> Date: 28 Jun 90 15:50:42 GMT Sender: daemon@ucbvax.BERKELEY.EDU Organization: The Internet Lines: 70 X-Unparsable-Date: Thu, 28 Jun 90 17:17:44 SWI I would like to thank the people who responded to my first message about our TCP/IP problems. However none of the solutions suggested have worked. Hopefully with more information someone can point me to a solution. We have an application that opens ~8 sockets to a host to do some handshaking. This host is in a VMS crate and likes to hang. So the operator then re-boots the VME. This is leaving the TCP sessions on our DN10Ks hung, as evidenced by this tcpstat. $ tcpstat Active connections Proto Recv-Q Send-Q Local Address Foreign Address (state) tcp 0 0 apomc10k.rlb o-oppbfe3.opdaq_or ESTABLISHED tcp 0 0 apomc10k.1259 o-oppbfe3.opdaq_or ESTABLISHED tcp 0 0 apomc10k.1258 o-oppbfe3.opdaq_or ESTABLISHED tcp 0 0 apomc10k.1257 o-oppbfe3.opdaq_or ESTABLISHED tcp 0 0 apomc10k.1256 o-oppbfe3.opdaq_pr ESTABLISHED tcp 0 0 apomc10k.1255 o-oppbfe3.opdaq_st ESTABLISHED tcp 0 0 apomc10k.1254 o-oppbfe3.opdaq_em ESTABLISHED tcp 0 0 apomc10k.1233 o-oppbfe3.telnet ESTABLISHED tcp 0 0 apomc10k.1188 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1187 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1185 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1184 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1183 o-oppbfe3.opdaq_pr FIN_WAIT_2 tcp 0 0 apomc10k.1176 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1158 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1157 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1155 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1154 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1153 o-oppbfe3.opdaq_pr FIN_WAIT_2 tcp 0 0 apomc10k.1081 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1080 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1079 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1078 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1069 o-oppbfe3.opdaq_pr FIN_WAIT_2 tcp 0 0 apomc10k.1044 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1043 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1042 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1039 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1037 o-oppbfe3.opdaq_pr FIN_WAIT_2 tcp 0 0 apomc10k.1033 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1032 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1030 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1029 o-oppbfe3.opdaq_or CLOSE_WAIT tcp 0 0 apomc10k.1027 o-oppbfe3.opdaq_or FIN_WAIT_2 tcp 0 0 apomc10k.1026 o-oppbfe3.opdaq_pr FIN_WAIT_2 And at this time here is the output to llkob | grep dev on the node R nR_xor_1W 1F3AD 1F3AD /sys/node_data/dev/lib/vxa_dev_int.lib R nR_xor_1W 1F3AD 1F3AD /sys/node_data/dev/lib/vxa_dev.lib W Cowriters 25AEB 1F3AD //op10k1/sys/node_data/dev/crp00 W Cowriters 25AEB 1F3AD //op10k1/sys/node_data/dev/crp00 So now the question is. How can I go about killing all of these FIN_WAIT_2 sessions? ( Again not as root, the sleep problem again ). I remember that there was a way to do this under SR9.7. Any help would be appreciated, as after many re-boots of the VMS side, we become unable to make a TCP connection. Ciao, Casper Paul Schenk | University of Victoria | CERN EP/ OPAL pschenk%cernapo@cernvax.cern.ch <- Pref. pschenk@cernvm.cern.ch schenk@uvvm.bitnet