Paul Hinker wrote:
About 3-4 weeks ago I suddenly started seeing high latency on the e1000g
interface of my Ultra 27.
ping -s yahoo.com
PING yahoo.com: 56 data bytes
64 bytes from ir1.fp.vip.re1.yahoo.com (69.147.125.65): icmp_seq=0. time=74.611
ms
64 bytes from ir1.fp.vip.re1.yahoo.com (69.147.125.65): icmp_seq=1.
time=5052.370 ms
64 bytes from ir1.fp.vip.re1.yahoo.com (69.147.125.65): icmp_seq=2.
time=4052.442 ms
64 bytes from ir1.fp.vip.re1.yahoo.com (69.147.125.65): icmp_seq=3.
time=3052.408 ms
64 bytes from ir1.fp.vip.re1.yahoo.com (69.147.125.65): icmp_seq=4.
time=2052.412 ms
64 bytes from ir1.fp.vip.re1.yahoo.com (69.147.125.65): icmp_seq=5.
time=1052.437 ms
In my experience, this type of readout indicates a problem with name resolution,
not necessarily a 'real' network latency. The times decrement by ~1s which is
ping's default packet interval. The responses queue up waiting for the reverse
lookup to complete and are then processed sequentially. In the above output, it
seems that it took ~5s to do the reverse lookup - suspiciously the same as the
default RES_TIMEOUT in resolv.h?
Use snoop/wireshark to confirm if the packet latency is real or not.
--
Jon Anderson
ORACLE RPE Systems
SPARC House (UK)
Tel: ++44 (0)1252 421 868
Mob: ++44 (0)7747 180 910
_______________________________________________
networking-discuss mailing list
[email protected]
