[ BCC [email protected] ]
OK. First point - Solaris 10 doesn't belong in this forum (which is
OpenSolaris related).
That said, I'm looking at a possible (elusive) bug in the gathering of
statistics by netstat in S10. I knew it affected tcp connections, but
was unsure how much of the netstat "subsystem" was affected.
<shameless hijack>
If anybody else reading this has any strange netstat outputs
(particularly the output from "netstat -an -P tcp -f inet"), I'd
appreciate knowing which kernel you were running (i.e. the output from
"uname -v" or "uname -a" would suffice).
</shameless hijack>
Can you mail me directly with any observations you have (such as the
uname output from above) and when it first started, how long this
application has been gathering statistics this way, etc...
I can't promise a solution for your issue, but if there is something
more fundamental wrong with netstat, I may end up fixing your issue as a
side-effect of the bug I'm fixing:
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6791831
Regards,
Brian
Zhiyuan Xu wrote:
It's Solaris 10.
We are sampling every 5 seconds to calculate the bandwith. 'netstat -s -P tcp 5' is called in a perl script with a pipe, when netstat print output every 5 seconds, perl code will read it and parse the tcpInInorderBytes we want. Then the script will go back to netstat again.
The large bandwidth value appeared exactly in 502 seconds interval.
You are right, it's 10 seconds from last number when that large bandwidth value appeared. There must be an abnormal netstat output. Thanks. I'll write that into log and get back here again.
Really appreciate your help.
--
Brian Ruthven
Solaris Revenue Product Engineering
Sun Microsystems UK
Sparc House, Guillemont Park, Camberley, GU17 9QG
_______________________________________________
networking-discuss mailing list
[email protected]
