RLOGIN fails to connect

[Wim Van den Wyngaert]

When I do a rlogin from VMS to VMS (all 7.3) I get rarely (once a month) and immediatly (within 1 sec) :

$ rlogin X
%RLOGIN-E-INETERROR, Internet interface error
-RLOGIN-I-INETCALL, connect()
-SYSTEM-F-TIMEOUT, device timeout

If I redo the operation it ALWAYS works.

I vaguely remember having this on 6.21H3 too.

Someone knows whats wrong ?

Wim

[labadie_1]

This reminds me of something. Which Tcpip version on both nodes ? Any message in operator.log or anywhere (rlogin log files) on the remote node ?

[Wim Van den Wyngaert]

No message on either node. The service rlogin has error, logi, logo, mdfy, rjct and timo as log option.

Nothing in accounting and audit.

TCP 5.3 eco 2 with rlogin version 5.3-18.

Wim

[Thomas Ritter]

May we see the output from

$ ucx show service rlogin/full

on the node which will be accepting the rlogin request ?

[Volker Halle]

Wim,

I'm guessing that you use the IP name for 'X'. So it could also have something to do with name to address translation.

Does it also happen, if you use the IP address (then we could rule out DNS) ?

Volker.

[Wim Van den Wyngaert]

We use the local host file only. No DNS.
Service description will folloz tomorrow.

Wim

[Wim Van den Wyngaert]

It folloWs now :
SPVMX2/MGRWVW>ucx sho servi rlogin/full

Service: RLOGIN
State: Enabled
Port: 513 Protocol: TCP Address: 0.0.0.0
Inactivity: 1 User_name: not defined Process: not defined
Limit: 50 Active: 4 Peak: 4

File: not defined
Flags: Case Listen Proxy Rlog Rtty IPv6

Socket Opts: Keepalive Rcheck Scheck
Receive: 3000 Send: 3000

Log Opts: Error Logi Logo Mdfy Rjct TimO
File: not defined

Security
Reject msg: TCPIP RLOGIN Connection refused
Accept host: 0.0.0.0
Accept netw: 0.0.0.0

[Wim Van den Wyngaert]

Getting closer.

When I do a rlogin to a node that is down, I get a device timeout too but preceded by "failed to connect to remote host".
Noticed with tcptrace that the arp request is sent once and the timeout is given after 1-2 seconds.

Didn't find a place where this timeout is configured. And I would sure like a retransmit in case of a packet loss.

So, if the node is not responding fast enough, I get the timeout ?

Found also that in 7.3 the protocol arp config is not in help but it exists (but not my search items).

Wim

[Wim Van den Wyngaert]

When a onnection is opened to a node that is down, 1 arp every second will be sent. If the answer doesn't arrive within tcp_keepinit (seconds or half seconds ?) the connection timeouts.

My tcp_keepinit is on 5 (seconds, or is it 2.5 seconds ?).

I measured the rlogin timeout again and it is after 1.8/2.9/1.7/1.8/2.0 seconds (waited some minutes between the trials).

Some rounding problem ?

BTW : the node from the original question is about 5 km from my starting node.

Wim

[Wim Van den Wyngaert]

Increased keepinit to 12 and now it takes about 5 seconds (as I initialy intended but the unit seems to be 0.5 sec and you loose maximum 1 sec in the roundings).

Now wait and see.

Wim

[Thomas Ritter]

Wim, at this stage I can only suggest you try and understand the network topology between the rlogin source and its destination. I guess it's not a single segment network. Prehaps there are some firewalls in between ? If so ask the network manager if the could monitor a connection request. If its a popular checkpoint firewall there will be a GUI which can provide a realtime view of all connection requests. Also is rlogin popular or is it just your connection choice ? Are there alternative ?

[Thomas Ritter]

From our working systems

$ ucx show ver

HP TCP/IP Services for OpenVMS Alpha Version V5.4 - ECO 5
on a COMPAQ AlphaServer DS20E 666 MHz running OpenVMS V7.3-2

$ ucx show service rlogin/full

Service: RLOGIN
State: Enabled
Port: 513 Protocol: TCP Address: 0.0.0.0
Inactivity: 1 User_name: not defined Process: not defined
Limit: 50 Active: 0 Peak: 1

File: not defined
Flags: Listen Rlog Rtty IPv6

Socket Opts: Keepalive Rcheck Scheck
Receive: 3000 Send: 3000

Log Opts: Acpt Actv Dactv Conn Error Logi Logo Mdfy Rjct TimO
File: not defined

Security
Reject msg: UCX RLOGIN Connection refused
Accept host: 0.0.0.0
Accept netw: 0.0.0.0
$

[Wim Van den Wyngaert]

Thomas,

Rlogin is the only service to login using a proxy. That's why I use it.
There are no firewalls but the route between the 2 nodes has 3 routers between them (en some switches). Speed is variable (from=10Mbit, netw=100Mbit, backbone=1Gbit, netw=100Mbit, to=100Mbit).

I guess the routers/destination node sometimes have something better to do and recat a bit slow.

I can not ask to monitor it because it only happens once a month and this is a 10000+ users network.

Thanks anyway.

Wim

[Wim Van den Wyngaert]

Problem still not solved. Got it again within 2 seconds while my timeout is now on 6 sec (tcp_keepinit=12).

Wim

[Wim Van den Wyngaert]

Did a rsh on this node to all other nodes (100). No problem.
Did 5000 rlogins to the node from the problem description. No problem.

Wim

[Wim Van den Wyngaert]

Did 5000 rlogins and cleared arp of the router between each rlogin. No problem.

Wim