HP3000-L Archives

August 2000, Week 2

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
"HOFMEISTER,JAMES (HP-USA,ex1)" <[log in to unmask]>
Reply To:
HOFMEISTER,JAMES (HP-USA,ex1)
Date:
Mon, 14 Aug 2000 13:41:01 -0600
Content-Type:
text/plain
Parts/Attachments:
text/plain (243 lines)
Hello Folks @ 3000-l,

Re: Telnet logon difficulties

In response to questions raised on 3000-L concerning increasing the
priority of the INETD process to the BS queue:

Here is what the system looks like today with the HP recommended
configuration of
the JINETD job with :run jinetd.net.sys;pri=cs

----------------------------------------------------------------------
:showproc ;pin=1;system;tree
QPRI  CPUTIME   STATE  JOBNUM  PIN  (PROGRAM) STEP
...
B149* 0:05.153  WAIT               24   (NMFILE.PUB.SYS)
B149* 0:00.510  WAIT               25   (NMLOGMON.PUB.SYS)
B148* 0:00.048  WAIT                 26   (NMLOGICS.PUB.SYS)
B149* 0:00.335  WAIT                 27   (NMTRCMON.PUB.SYS)
B149* 1:55.567  WAIT                 28   (NMCONSOL.PUB.SYS)
B149* 0:03.462  WAIT               34   (NETCP.NET.SYS)
C152* 0:00.038  WAIT                 52   (ICMPSERV.NET.SYS)
B149* 0:00.276  WAIT                 53   (SOCKREG.NET.SYS)
B149* 0:00.033  WAIT                 54   (PT2PNSTN.NET.SYS)
B149* 0:00.837  WAIT               51   (TCPSIP.NET.SYS)
B149* 0:03.678  WAIT               56   (DSDAD.NET.SYS)
...

B152  0:00.456  WAIT   J2          48   (JSMAIN.PUB.SYS)
D202  0:00.439  WAIT   J2            57   :RUN inetd.net.sys;pri=cs
C152  0:00.965  WAIT   J2              59   (INETD.NET.SYS)
...
C152* 0:52.714  WAIT             46   (SNMP.NET.SYS)

Bootpd request:

C200  3:03.110  WAIT   J2                39   (BOOTPD.NET.SYS)

Telnet logon:

B152  0:00.452  WAIT   S9          66   (JSMAIN.PUB.SYS)
C152  0:00.161  WAIT   S9            45   :EDITOR
C152  0:00.465  WAIT   S9              39   (EDITOR.PUB.SYS)

VT logon:

B152  0:00.441  WAIT   S8          42   (JSMAIN.PUB.SYS)
C152  0:00.709  WAIT   S8            67   :EDITOR
C152  0:00.325  WAIT   S8              64   (EDITOR.PUB.SYS)

FTP logon:
C152  0:00.901  WAIT   J28               64   (FTPSRVR.ARPA.SYS)

Note: As per the above bootpd request, the priority of this process
has already dropped due to incoming request being processed.
----------------------------------------------------------------------

In trouble shooting situations (not production) I have implemented a
PRI=152 to lock INETD in a linear queue at the top of the CQ with
priority = 152 and here is what that looks like:

----------------------------------------------------------------------
B152  0:00.466  WAIT   J26         43   (JSMAIN.PUB.SYS)
D202  0:00.404  WAIT   J26           57   :RUN inetd.net.sys;pri=152
B152  0:00.917  WAIT   J26             61   (INETD.NET.SYS)

Bootpd request:

B152  0:00.634  WAIT   J26               64   (BOOTPD.NET.SYS)

FTP logon:

B152  0:00.695  WAIT   J26               63   (FTPSRVR.ARPA.SYS)

Telnet logon:

B152  0:00.436  WAIT   S10         66   (JSMAIN.PUB.SYS)
C152  0:00.187  WAIT   S10           39   :EDITOR
C152  0:00.157  WAIT   S10             45   (EDITOR.PUB.SYS)

I have not checked to see what happens to SAMBA and all of the
other servers running under INETD, but I do see an undesirable result
here with FTP and BOOTPD running at linear queue B152.
----------------------------------------------------------------------

Now as per the suggested configuration of ALTPROC JOB=#J5;PRI=BS.

I did this on my system:

----------------------------------------------------------------------
:showjob job=jinetd,manager.sys

JOBNUM  STATE IPRI JIN  JLIST    INTRODUCED  JOB NAME

#J26    EXEC        10S LP       MON 12:34P  JINETD,MANAGER.SYS

JOBFENCE= 0; JLIMIT= 60; SLIMIT= 100

:altproc job=#j26;pri=BS

B152  0:00.466  WAIT   J26         43   (JSMAIN.PUB.SYS)
B100  0:00.404  WAIT   J26           57   :RUN inetd.net.sys;pri=152
B152  0:01.219  WAIT   J26             61   (INETD.NET.SYS)

Telnet logon:

B152  0:00.440  WAIT   S11         66   (JSMAIN.PUB.SYS)
C152  0:00.279  WAIT   S11           39   :EDITOR
C152  0:00.269  WAIT   S11             59   (EDITOR.PUB.SYS)

Bootpd request:

B152  0:01.340  WAIT   J26               64   (BOOTPD.NET.SYS)

FTP logon:

B152  0:00.654  WAIT   J26               63   (FTPSRVR.ARPA.SYS)

Again I have not checked to see what happens to SAMBA and all of the
other servers running under INETD, but I do see an undesirable result
here with FTP and BOOTPD running at linear queue B152.
----------------------------------------------------------------------

One of my colleagues (Jay Bauer) saw a different result so I tested
further and found if the JINETD job does not already have the HP
recommended ;PRI=CS on the run that the result of the ALTPROC is
different:

----------------------------------------------------------------------
B152  0:00.468  WAIT   J27         48   (JSMAIN.PUB.SYS)
B100  0:00.403  WAIT   J27           59   :RUN inetd.net.sys
B100  0:00.894  WAIT   J27             64   (INETD.NET.SYS)

Bootpd request:

B100  0:00.643  WAIT   J27               67   (BOOTPD.NET.SYS)

FTP logon:

B100  0:00.651  WAIT   J27               45   (FTPSRVR.ARPA.SYS)

Telnet logon:

B152  0:00.430  WAIT   S13         66   (JSMAIN.PUB.SYS)
C152  0:00.246  WAIT   S13           39   :EDITOR
C152  0:00.156  WAIT   S13             61   (EDITOR.PUB.SYS)

Again I have not checked to see what happens to SAMBA and all of the
other servers running under INETD, but I do see an undesirable result
here with FTP and BOOTPD running at linear queue B100.
----------------------------------------------------------------------


Concerns after looking into all of the above data...

First of all I am assuming your system queues are something close to
the defaults...

                    ------QUANTUM-------
QUEUE  BASE  LIMIT  MIN    MAX    ACTUAL  BOOST  TIMESLICE
-----  ----  -----  ---    ---    ------  -----  ---------
 CQ    152    200   1      2000   30      DECAY    200
 DQ    202    238   2000   2000   2000    DECAY    200
 EQ    240    253   2000   2000   2000    DECAY    200

Note:  Actually when I was a sysmgr (In another life) I always wrapped
the DQ slightly over part of the CQ to assure my batch jobs at lease got
logged on and a chance to execute.

Concerns:
1) Telnet is the only server (due to it's proprietary implementation)
which launches correctly in this ";pri=bs" scheme.  I am ok with the
JSMAIN at B152 (same as VTSERVER and DTC) and the CI and it's children
behave executing at C152 and decay to C200 as per the above noted queue
defaults.

2) FTP at B152 is a problem.  Process in the BQ don't decay!  They
also don't block for lower priority processes.  With FTP at B152 on a
low to mid range or busy system and a client sending data from a high
end system, FTP at B152 will degrade the performance of online sessions
Telnet and NS-VT.  ICMPSERV (ping and ICMP message server) and SNMP
Servers will time out request.

3) FTP at B100 is a serious problem.  At the very top of this document
I perform a "showproc ;pin=1;system;tree" and I identify the network
system process's.  Note that all of these process are running at B149
except for ICMPSERV and SMMP which run at C152.  I could ramble on
about possible concerns with semaphore locking, hangs, looping, but
the bottom line is a System Abort is eminent.

4) SAMBA and all of the other servers running under INETD.  Pretty much
the same concerns as above for FTP at both the B152 and B100 priorities.
An added concern is with INETD at a 'system process' priority and with
the child process's being created at these same levels is that INETD is
a standard (I have a definition of this I will give you later) and that
increases the likely hood that an external or internal hacker who has
the ability to launch INETD service request on your system will more
likely be successful at launching a loss of service attack which in this
case will not only take out the service, but may also hang our system.

My definition of a "standard" is mediocre "the minimal functionality
which satisfies the fewest people contracted to be built by the lowest
bidder." //:~)

The feature set we are looking for is for INETD to create process and
services at the correct priorities and assure sufficient resources.
This is a feature of "DSDAD" HP's proprietary (ops, said that bad word)
NS Services Server.  INETD does not have this functionality.  DSDAD as
you can see in the above "showproc ;pin=1;system;tree" executes at B149
which assures it will never be starved for resources, but it creates
the servers at appropriate priorities which are adjusted by HP Labs to
assure sufficient resources to perform the required task, but avoiding
interference with other system processes and unfair competition with
user processes.

Bottom Line:
------------

I would not recommend the PRI= be changed from the PRI=CS recommended
by HP.

I would suggest you pursue an enhancement request for INETD to increase
the functionality to support running INETD as a system process and
create the service processes at the appropriate user priority levels.
I have recommended this solution in the past SR 4701-371708 but I was
disappointed with the final ";PRI=CS" solution.  I would suggest a new
SR - Enhancement Request be submitted and of course talk up this need
at HP-World.

Note: HP does not support increasing the priority of INETD or any user
applications to a priority in the BS queue.

FYI: Before you assume you are seeing a performance problem with INETD
launching services, please make sure you have the General Release FTP
and Telnet patches installed as both contain fixes for hangs of INETD.

Regards,

James Hofmeister
Hewlett Packard
Worldwide Technology Network Expert Center
P.S. My Ideals are my own, not necessarily my employers.

ATOM RSS1 RSS2