TCPIP ftp problems

[tanja_6]

Hi!
I'm new here.
I am running OpenVMS 7.2-1 with TCPIP Services version 5.0a ECO 3 on a Alphaserver 1000 2/266

The problem is that ftp is behaving inconsistently.
Some time it works OK, some time it needs long time to connect or work.
When it works slowly, same behavior shows when trying ftp locally on server.
Telnet works OK whole time.

For example:

$ ftp tango
220 tango.tkczag.ht.hr FTP Server (Version 5.0) Ready.
--- waiting 30 second
Connected to TANGO.TKCZAG.HT.HR.
----- waiting 67 seconds
Name (TANGO.TKCZAG.HT.HR:tanja):
331 Username tanja requires a Password
--- waiting 28 seconds
Password:
------------------------------------

$ TCPIP show dev

shows:
bg1052 STREAM 21 0 FTP 0.0.0.0
bg1207 STREAM 21 3629 FTP TANJA_PC
bg1229 STREAM 21 2808 FTP 192.168.28.78
bg1239 STREAM 21 42085 FTP 192.168.115.103
bg1251 STREAM 21 56811 FTP 172.31.27.220
bg1258 DGRAM 1230 0 0.0.0.0
bg6035 STREAM 21 38308 FTP 192.168.80.69
bg6072 STREAM 20 38367 192.168.80.69
bg8928 STREAM 0 0 0.0.0.0
bg9164 STREAM 0 0 0.0.0.0

Devices bg9164, bg 8928 belong to ftp proces:
$ ucx show dev bg9164/ful
Device_socket: bg9164
Type: STREAM LOCAL REMOTE
Port: 0 0
Host: 0.0.0.0 0.0.0.0
Service:

RECEIVE SEND
Queued I/O 0 0
Socket buffer bytes 0 22
Socket buffer quota 33580 33580
Total buffer alloc 0 512
Total buffer limit 268640 268640
Buffer or I/O waits 0 0
Buffer or I/O drops 0 0
I/O completed 3 3
Bytes transferred 28 120

Q0LEN 0
QLEN 0
QLIMIT 0
TIMEO 0
ERROR 54
OOBMARK 0



Options: REUSEADR OOBINLINE
State: CANTSENDMORE CANTRECVMORE PRIV
RCV Buff: None
SND Buff: None


All help appreciated.

Regards
Tanja

[Karl Rohwedder]

Just a guess, maybe the FTP server process has timed out. So when receiving a new connection request, the INETD has to start a new FTP server process (1st wait). The FTP server process then starts a child process to handle your transfer (2nd wait).

If I remember correctly, the FTP server process can be told to not time out (via a logical name?).

You may also have a look into your systemwide loginfiles, if on network mode the normal interactive things are skipped (IF F$MODE().eqs. "NETWORK" THEN $EXIT)

regards Kalle

[Volker Halle]

Tanja,

welcome to the OpenVMS ITRC forum.

FTP can be sensitive to DNS lookup problems.

Does UCX SHOW NAME list multiple BIND servers ? Do they work o.k., when you see your FTP problem ? You could test this with TCPIP$NSLOOKUP.

Volker.

[Robert Gezelter]

Tanja,

Also check your network infrastructure. Telnet produces effectivly zero load on the network.I have seen situations where infrastructure problems (e.g., full/half duplex mismatches) did not show up when using telnet (or DECnet SET HOST), but did show up severly whenusing FTP or DECnet COPYs.

-Bob Gezelter, http://www.rlgsc.com

[tanja_6]

This server is used as ftp server in our environment for 5 years on this software level, and severe problems with ftp are reported last month.
My network guys claim that problem is on server which fits with problems with local ftp on server. There are times when ftp works fast for hours, but this week we have big problems.

the main thing that I observed is:

if ftp is not working properly, there exists device like
bg2399 DGRAM 2139 0 0.0.0.0
where device number grows and local port number grows, as if something is scanning ports.
When there are no such devices, ftp work OK.

[Volker Halle]

Tanja,

if there is some node doing port scans on your FTP server, this could certainly explain your FTP performance problems. TCPIP$INETACP may be busy creating and deleting those BG devices.

Ask your network guys to trace IP network traffic to your node and find out, where those port scans are coming from.

Volker.