[Samba] Thousands of TIME_WAIT connections with CUPS?
esiotrot at gmail.com
Thu Nov 5 00:13:54 MST 2009
2009/11/5 Adam Nielsen <adam.nielsen at uq.edu.au>:
>> I don't have an answer for your actual question, but if it takes
>> several seconds to get this list, it's probably because of DNS
>> queries, and using "-n" will speed it up dramatically.
> Thanks for the suggestion. I was actually using "cat /proc/net/tcp" as
> this is what the monitoring software uses, so there's no DNS lookups
> there. Just the sheer number of them (and any processing done by the
> monitoring software) is the cause of the delay.
Ah, OK. Then is it possible to get your monitoring software to ignore
the 127.0.0.1:631 connections?
> Adding -n when using 'netstat' only shaves off 0.3 seconds, I guess
> because 99% of them are all "localhost"!
> But of course the connections shouldn't be there in the first place, so
> any ideas how to get rid of them would be much appreciated! I am only
> running Samba 3.2 though, so perhaps the problem will go away when I get
> a chance to upgrade to 3.4...
I don't think that will help.
See the following:
The diagram there is a bit corrupted from the original ASCII art
version (missing sequence numbers etc.), so see figure 13 from here:
Michael Wood <esiotrot at gmail.com>
More information about the samba