On Saturday 24 January 2004 12:35 pm, John Andersen wrote:
On Saturday 24 January 2004 08:43, Robert Fitzpatrick wrote:
I just started noticing a problem ... Logins with KDE take a minute or longer. When I ping machines by name at this particular data center, there are several seconds between responses, but replies are 100%. I try the same ping by name from a local Windows server and no delay at all. [snip]
Paragraphs were invented for a reason. Please use them.
without trying to be snide, I'll add a "ditto" to that -- you ended up writing so much people tend to focus on the last part and may miss the important point you made at the start: THERE ARE SEVERAL SECONDS BETWEEN RESPONSES for the linux machine, but not the windows machine. I think [but could be wrong] that this effectively eliminates any "DNS" problems because I don't believe there is a DNS lookup FOR EACH ping. The general rule-of-thumb is that you evaluate the user input, and if it is a direct IP address you use it; if not, you perform one lookup and remember that for later connection attempts, but I'm digressing If you had said "there is a pause for several seconds before the first ping returns, but then it proceeds normally" I'd be inclined to follow the DNS path. but with several seconds between EACH ping, I'd look at the hardware [such as the switch] and I'd start checking to see if perhaps another machine is configured with the same IP or MAC address as your linux machine [same IP can be pretty common, same MAC is rather difficult, but not impossible] network analyzers like ethereal are good for tracking down these types of problems. check cabling -- try swapping the cable/port used by the windows machine for the one on the linux machine and see if the problem follows the cable or the port Check to see if you are getting "errors" on your network interface -- the command "ifconfig" will report this as follows: bigbro:~ # ifconfig eth0 Link encap:Ethernet HWaddr 00:40:F4:6E:4D:22 inet addr:192.168.40.2 Bcast:192.168.40.255 Mask:255.255.255.0 inet6 addr: fe80::240:f4ff:fe6e:4d22/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:402169 errors:0 dropped:0 overruns:0 frame:0 TX packets:372014 errors:0 dropped:0 overruns:5 carrier:0 collisions:0 RX bytes:171030817 (163.1 Mb) TX bytes:56593843 (53.9 Mb) Those last three lines are what you are looking for -- heavy errors are never good. Finally [and this is a bit of a stretch], how are the interrupts on your system? Look at the pseudo-file /proc/interrupts and look for your network device. do "some" activity, and look again -- do the numbers seem reasonable for the amount of traffic generated? Some of the >ahem< low-end network cards are "win-lan" cards [think "winmodem"] and rely heavily on the CPU -- I have one that would regularly crash my 100mhz file server, but worked fine in my 466mhz "client" computer
1st you should compare the dns servers used by the fast windows box and the slow linux box and see if they are IN FACT the same.
Actually, the first thought that came to mind [back on the DNS front] was to make sure that the windows machine doesn't have it's address cached locally or in a local "hosts" file. -- Yet another Blog: http://osnut.homelinux.net