tftpd and timeout

[Mark Greene_1]

I am running tftp on a D-class server running 10.20 and am getting the message "Timeout (no requests in 10 minutes)" in syslogd. It appears that tftpd runs for 10-15 minutes, dies, and anywhere from 10-20 minutes later stats up and the cycle begins again.

We are using the system to store init files for about a dozen terminal servers, and if tftpd is not running when a term server gets reset, the term server hangs and we have to power it off/on until it connects with the system to get its start-up files.

I've looked in the man pages for tftp and inetd and did not see any options to change this. In the tftp man page it states there is are -R and -T options for retransmission timeouts, but the default values are 5 and 25 seconds and neither are specified in inetd.conf, so I doubt that's it.

I've also done search on hp's site and turned up nothing. Any suggestions on where else to look are appreciated.

--
mark

[Uday_S_Ankolekar]

Hi,
Look at this rfc for TFTP Timeout Interval and Transfer Size Options.

http://www.faqs.org/rfcs/rfc2349.html

Also if you know if there is a header file, "tftp.h", the values for TFTP_RETRY and
TFTP_TIMEOUT are set to 10 and 144 (8 seconds) respectively try with those settings


hope this helps.

-USA..

[Craig Rants · Accepted solution]

I guess I don't understand. If tftp is in inetd, then when inetd gets the tftp requst it will spawn tftpd to handle the request. What do you mean when you say the terminal server gets reset? Maybe you need to increase the timeout of the terminal server so the tftpd has enough time to spawn and answer the request.

Maybe this will help,
C

[Bill Thorsteinson]

I would expect tftpd to timeout after a while with no requests. inetd should be able to start tftpd fast enough. The following line is from our linux box that is used to load a cisco router.

tftp dgram udp wait nobody /usr/sbin/tcpd in.tftpd /var/cisco

This should invoke the tftp daemon when needed. It will then time out some time after all the files have been loaded.

[Mark Greene_1]

This is what I am seeing in syslogd:

Jan 8 13:21:28 xxx tftpd[20967]: Timeout (no requests in 10 minutes)
Jan 8 13:50:29 xxx tftpd[20974]: Timeout (no requests in 10 minutes)
Jan 8 14:02:37 xxx tftpd[20989]: Timeout (no requests in 10 minutes)
Jan 8 14:19:29 xxx tftpd[21039]: Timeout (no requests in 10 minutes)
Jan 8 14:48:29 xxx tftpd[21056]: Timeout (no requests in 10 minutes)
Jan 8 15:17:29 xxx tftpd[21060]: Timeout (no requests in 10 minutes)
Jan 8 15:46:29 xxx tftpd[21067]: Timeout (no requests in 10 minutes)
Jan 8 16:15:29 xxx tftpd[21073]: Timeout (no requests in 10 minutes)
Jan 8 16:44:29 xxx tftpd[21084]: Timeout (no requests in 10 minutes)
Jan 8 17:13:29 xxx tftpd[21089]: Timeout (no requests in 10 minutes)

If you guys are saying that this is normal behavior, that's cool. I just could find anything definitive that stated tftpd ran only on demand and was not persisent. I thought that might be the case given that it runs under inetd, but was not sure.

I have had to explicitly start tftpd to get the termservers to be able to init on power-up, and that's that part that worries me. I suppose I will have to keep an eye on it closer, and log the time when we cycle the term servers and compare that with the tftpd times in the syslog. The problem there is that I don't get a message when tftp starts. I am hoping if I start inetd with the -l, that inetd will log that bit for me.

Thanks to all for the responses.

mark