From: Greg Gallagher (sun-managers@dahak.com)
Date: Wed Jan 08 2003 - 12:23:18 EST
Hello fellow Sun Managers,
I'm seeing a consistent problem with our servers running TIBCO's
Rendezvous product (rvd/rvrd) on Solaris servers with a statistic in
netstat called "udpInOverflows".
Here's what I see running "netstat -s":
IPv4 ipForwarding = 2 ipDefaultTTL = 255
ipInReceives =411135321 ipInHdrErrors = 0
ipInAddrErrors = 0 ipInCksumErrs = 0
ipForwDatagrams = 0 ipForwProhibits = 0
ipInUnknownProtos = 5100 ipInDiscards = 0
ipInDelivers =495773971 ipOutRequests =132672232
ipOutDiscards = 0 ipOutNoRoutes = 0
ipReasmTimeout = 60 ipReasmReqds = 0
ipReasmOKs = 0 ipReasmFails = 0
ipReasmDuplicates = 0 ipReasmPartDups = 0
ipFragOKs = 0 ipFragFails = 0
ipFragCreates = 0 ipRoutingDiscards = 0
tcpInErrs = 0 udpNoPorts =1104118
udpInCksumErrs = 0 udpInOverflows = 52626
rawipInOverflows = 0 ipsecInSucceeded = 0
ipsecInFailed = 0 ipInIPv6 = 0
ipOutIPv6 = 0 ipOutSwitchIPv6 = 2259
On some servers, the udpInOverflows ticks up, mostly during high
periods of activity but not consistently, and there is no set
udpInDatagrams delta that I can see which collates to this value
(i.e. if I have high traffic on my TIB bus, I see the value tick up
occasionally if I'm pushing 4000 datagrams / seconds across the bus,
or fewer, or sometimes none even when I'm pushing across more.
One additional note is that I see this problem happen less on
Solaris 8 and more on Solaris 2.6 boxes... hrm.
The only thing I understand about the "udpInOverflows" value in
"netstat -s" is that it means a remote udp socket's receive buffer
wasn't emptied when trying to send data. At least that's what I
think.
The correlation of TIBCO with this udpInOverflows isn't proven, but
I'm curious what other people are seeing, especially if they're
running TIBCO, and how can I fix this (perhspa a /dev/ip or /dev/udp
parameter change?). Especially if it's a network performance issue, a
la re transmitted tcp packet rate, clients or servers are basically
dropping udp packets which the TIBCO daemon must re-transmit.
Thanks in advance for any information or help,
regards,
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:25:33 EDT