Alexander Hall wrote:
Dec 15 01:30:50 vtftp1 tftpd[5866]: recv: Connection refused
# zcat daemon.2.gz | grep 'Connection refused'
# zcat daemon.1.gz | grep 'Connection refused'
# zcat daemon.0.gz | grep 'Connection refused'
# cat daemon | grep 'Connection refused'
Oh. What you see seems to be caused by failing system calls, not any
userland inetd limit.
No idea why.
/Alexander
Why would you say that? May be true, I didn't look in the direction
before as doing what I said address the problem, but that may be a good
suggestion however. I assume the 'Connection refused' was for limits
impose on tftpd, from inetd as I have lots of successful connections in
that same time frame, but as the log show, I sure was exceeding the 256
per minutes for sure in this case, so that limit was a logical place to
look at for me.
More from logs, just as details in two 10 seconds period:
# zcat daemon.3.gz | grep 'Dec 15 01:30:2' | wc -l
90
# zcat daemon.3.gz | grep 'Dec 15 01:30:3' | wc -l
92
Got this number of refuse connections in that same window, the others
were successful when they were retried a few seconds later. Just
couldn't handle the peek:
# zcat daemon.3.gz | grep 'Dec 15 01:30:2' | grep 'Connection refused' |
wc -l
6
# zcat daemon.3.gz | grep 'Dec 15 01:30:3' | grep 'Connection refused' |
wc -l
42
You can see the increase number of deny connection in that second 10
seconds window. This quickly get itself all worked out as the
connections are retried in the seconds that follow and all get served,
but it can't handle the peek as show above.
Interesting isn't it?
Anyway, that's what I try to resolve, so any good idea are welcome, but
that's where I am.
