AIX Telnet problems

[Scott Howell]

We have been having problems with our Unix server.
Every 3 or 4 days, the system begins refusing telnet connections
and the console login does not work. I can ping the server and ftp
appears to work, but I can not login or even establish a telnet or
Xwindows session. I cannot even login into the console.
I have to reboot the machine before things work
properly. The problem will reoccur about 3 or 4 days after the reboot. If
you have any idea of things to check, I would be glad to hear it. We are
running AIX 4.3.3 on an RS 6000 with the following characteristics:
2 375mhz processors
IBM H70
512 MB RAM
1 9.1 GB hard disk for OS
2 9.1 GB for data

Software running on this box
1 instance of Oracle 8.1.5.0.0
1 instance of Oracle 8.1.6.0.0
IBM WebSphere

This problem did not start until I added the second instance which was
Oracle 8.1.5.0.0. These Oracle instance are under different ORACLE_HOMEs. I
have run multiple Oracle version instances on a test platform before on an
HP D380 without problems. I don't think this is a problem, but I know it is
not recommended on a production box. FTP, seems to still work so it might be
hosing the telnetd? We can also still access Oracle through SQL*NET from a
PC client. Is there some kernel parameter we need to increase??

[Nicholas Dronen]

Scott Howell <scott...@mediaone.net> wrote:

> We have been having problems with our Unix server.
> Every 3 or 4 days, the system begins refusing telnet connections
> and the console login does not work.

What's the error? Do you get a login prompt at all?
What happens exactly?

Regards,

Nick Dronen

[Scott Howell]

The error I get from either the console or a telnet session from a PC is a
'Connection Refused'.
"Nicholas Dronen" <ndr...@io.frii.com> wrote in message
news:j_lS5.646$Bf7.19...@news.frii.net...

[Nicholas Dronen]

Scott Howell <scott...@mediaone.net> wrote:

> The error I get from either the console or a telnet session from a PC is a
> 'Connection Refused'.

I would be very surprised were you to get a 'connection refused'
error at the *console*. Nevertheless, it sounds like inetd
is dying. Can't you at least do a command-line login at the
console?

Regards,

Nick Dronen

[bob]

Okies, just ma 2c, check for:

1) named dying, telnet doesnt like this much
2) telnetd -a in the /etc/inetd.conf

if you find the -a, remove the -a
if you find named dying fixit. If it isn't running then patch
the unit.

what version?

brett

In article <j_lS5.646$Bf7.19...@news.frii.net>, ndr...@io.frii.com
says...

[Stefaan A Eeckels]

In article <bsnS5.653$Bf7.18...@news.frii.net>,

Given the fact it started after he added a second instance of
Oracle, he might be out of paging space. See the thread
"AIX 4.3.3 conks out under heavy load"
Message-ID: <8up728$1bf$1...@lure.pipex.net> in this group.

Take care,

--
Stefaan
--
Ninety-Ninety Rule of Project Schedules:
The first ninety percent of the task takes ninety percent of
the time, and the last ten percent takes the other ninety percent.

[Stephen Wylie]

We use telnetd without named running.
What does the -a mean? It's not a valid argument to the telnetd command.

Puzzled
Steve

bob wrote in message ...

[Tony Curtis]

>> On Tue, 21 Nov 2000 15:33:13 -0000,
>> "Stephen Wylie" <st...@cpcomputers.co.uk> said:

> We use telnetd without named running. What does the -a
> mean? It's not a valid argument to the telnetd command.

It's an argument to the kerberised telnetd (authentication
style), q.v.

hth
t
--
Eih bennek, eih blavek.

[Giovanni Vernucci]

Hi,

I think that you can check the mbuf value using the "netstat -m" command. Maybe
you have "memuse" value near to "memmax"
value.
In this case you have to change the "thewall" value using "no -o
thewall=<newvalue>".
You can also watch if there is a process which use large size of memory and it
grows in the time. You can use "ps aux | more"
Hope this is useful.

Gianni and Mario.

Please remove "forzaroma" from my email address to email me.

[bigtiny]

Are you allowed to take down one of the Oracle instances and see if the
problem persists?

This might help you to localize what's going on.

It's interesting that you can still get to the box via the Oracle-based
connectivity...

I wonder if you're tying up tcp/ip ports and not freeing them....?

k

Scott Howell <scott...@mediaone.net> wrote in message
news:b6kS5.1445$XX6.2...@typhoon.southeast.rr.com...

[bob]

the telnetd -a was supposed to speed up telnet connections if memory
serves me correct. All it did was bring our production box to it's
knees.

We moved all of the users over on a weekend and prepped for chaos
on the monday, no chaos.... by wed users were reporting connection
problems.

Apparently this is fixed in later versions

brett

In article <8ve4gl$557$1...@lure.pipex.net>, st...@cpcomputers.co.uk says...

> We use telnetd without named running.
> What does the -a mean? It's not a valid argument to the telnetd command.
>

> Puzzled
> Steve
>
> bob wrote in message ...

[Ejimofor]

I am currently experiencing very similar problems in my attempt to long
on to an RS/6000 F80 running AIX 4.3.3. It outputs an error message as
follows: "Telnet: Unable to connect to remote hosts: Connection
refused". Never experienced any such predicament on SUN/Solaris 2x. Any
clues?

Thank you,

Ejimofor Kenneth Nwoye.

[Janos]

Are the Oracle instances and WebSphera running with the same user - root? Is
it possible that one of them is forking some child processes and you are out of
available prcesses allowed for the user ?

[Scott Howell]

Thanks,
I will try this suggestion.
<Ejimofor>; "Kenneth Nwoye" <ek.n...@netdecisions.co.uk> wrote in message
news:20001122...@mis.configured.host...
Thanks for your reply. Infact I eventually sorted out the problem by
removing the -a flag from the telnetd entry in /etc/inetd.conf file. I
then executed the command 'startsrc -s inetd', in order to startup
inetd to allow telnet sessions. I have never encountered this problem
on Solaris 2x, in my experience.

Best wishes,

Mr Ejimofor Kenneth Nwoye,
Systems Administrator/Developer,
netdecisions (UK) Ltd,
Elsinore House, 77 Fulham Palace Road,
Hammersmith, London,
England, W6 8JA.
http://www.netdecisions.co.uk
Tel: +44 (0)208-222 9400 Ext.1226
Fax: +44 (0)208-222 9401
Mobile: +44 (0)7971 607 678
e-mail: ek.n...@netdecisions.co.uk

> On Wed, Nov 22, 2000 at 01:17:17PM +0000, Ejimofor wrote:
> > I am currently experiencing very similar problems in my attempt to
> > long on to an RS/6000 F80 running AIX 4.3.3. It outputs an error
> > message as follows: "Telnet: Unable to connect to remote hosts:
> > Connection refused". Never experienced any such predicament on
> > SUN/Solaris 2x. Any clues?

> Connection refused usually means no one is listening on the
> remote address:port. Check whether inetd is running. If
> it is, perhaps you're running out of mbufs (thought I doubt
> it). Check the error report for low mbuf errors. If you
> see them, double the thewall limit (max number of mbufs),

> # no -a thewall=[new value]

> Regards,

> Nick Dronen

[Norman Levin]

I have winnt on desktops, exceed, and I am getting the 'chooser' screen.
I select my AIX host and get the CDE login screen.

As root, I can log in to aix.
As an ordinary user, just 'bounce' back to cde login screen.
This occurs for a brand NEW user so there is no .Xauthority or really
any files initially - though after the attempt to come in thru X, you
do see these files in the 'team02' home directory.

Sounds like a permission problem but??? any ideas?

aix 4.3.3
--
Norman Levin

[Tom Hoffmann]

I have the same setup (Exceed/Aix 4.4.4) and have never seen this
problem. In thinking about it, how would CDE handle the forced password
change that would be required of a new user?

[Dan Jennings]

If the user's password has expired, an non-closable xterm starts and
prompts the user to set a new password.

Typically the reason a CDE login session bombs can be traced back
to the user's .profile or .dtprofile. If either of these contains
commands that need to interact with a terminal session (i.e. stty)
the CDE login will get to that point in the desktop login, catch
the terminal interaction and fail. Also, any errors that are
encountered during user login (again typically in the user's
.profile) will cause the CDE login to bomb out.

HTH,

Dan

[Norman Levin]

Dan Jennings wrote:
>
> If the user's password has expired, an non-closable xterm starts and
> prompts the user to set a new password.

This is for a BRAND NEW USER. Only thing in their home directory is
the .profile copied over by the mkuser procedure. Definitely has valid
password. So, while following paragraphs are all good info, don't think
it applies to my case.

>
> Typically the reason a CDE login session bombs can be traced back
> to the user's .profile or .dtprofile. If either of these contains
> commands that need to interact with a terminal session (i.e. stty)
> the CDE login will get to that point in the desktop login, catch
> the terminal interaction and fail. Also, any errors that are
> encountered during user login (again typically in the user's
> .profile) will cause the CDE login to bomb out.
>
> HTH,
>
> Dan
>

--
Norman Levin

[Tom Hoffmann]

On Fri, 01 Dec 2000 02:43:31 -0500, Norman Levin wrote:
>Dan Jennings wrote:
>>
>> If the user's password has expired, an non-closable xterm starts and
>> prompts the user to set a new password.
>
>This is for a BRAND NEW USER. Only thing in their home directory is
>the .profile copied over by the mkuser procedure. Definitely has valid
>password. So, while following paragraphs are all good info, don't think
>it applies to my case.

Any ~/.dt/errorlog or ~/.dt/startlog available?