Both with my current machine (a shiny new G5 dual 2.7 GHz machine with 4 GB of memory) and my previous G4 400 MHz (an antique), I have encountered a problem with the machine freezing up when there is a network access. I noticed a similar post when doing a search, but no useful replies. While the ultimate cause is unclear, the symptoms usually start with a network access (e.g. the web or mail). Safari seemed to be particularly good at causing the problem (hence I am now using Camino as a default browser). The symptoms basically are an application accesses the internet (say a mail server) and doesn't come back. Other applications seem to run normally, but no new applications can be started and network access results in a spinning beachball. I always have a terminal window open and while the shell is initially fine, attempting to issue a sudo reboot will result in a beachball as well. There is nothing notable in the system log after (a power button) restart. This is very frustrating as I need to run long calculations (the reason I got the machine) and this problem reduces reliablity to the extent it makes doing long calculations a risky proposition. I don't suspect that it is a hardware problem as the exact (symptom wise problem happened with the older machine). I have a pc and a G4 powerbook on the same 100 base T (cheapo) hub and these keep humming without problems (well windows still sucks but it is a not a network problem). I am now at the point where I am going to go out and buy a new hub (perhaps a firewall) to avoid funny packets that could be screwing up the ethernet device driver (if that is what it is), but I am curious how common the problem is and what may help.

I would use a router/switch in any event. There is an issue with OS X 10.3.5+ where you need to manually enter the network settings for primary and secondary DNS. I've also seen where a router needed an update. The cost of router/switch makes it easy and affordable, even all in one cable or DSL model with built-in router/switch.

I have a fixed IP address and I have entered primary and secondary DNS addresses that are responsive. I have no idea what the problem is, but am very curious if it is cause by ethernet signal noise that a firewall might block. I guess another question I have is would a router block strange packets as well?

Check your system log for lookupd segmentation fault or bus errors. Sudo hangs if lookupd is down. You can try sudo su before the system locks up, then killall lookupd or reboot.

For more info, search for lookupd in Apple Discussions > Safari

I read the posts regarding lookupd and they seemed relevant. I installed the daemon unlookupd, however, when the same problem occurred this morning, the system log just reported unlookupd was attempting to reset lookupd (hundreds of times). Lookupd appeared to be hung on exiting:
UID PID PPID CPU PRI NI VSZ RSS WCHAN STAT TT TIME F COMMAND
0 373 1 0 0 0 0 0 - Es ?? 0:00.00 6000 (lookupd)

I had a root shell running just to try out reboot and shutdown. They execute without locking up the terminal, but block:
UID PID PPID CPU PRI NI VSZ RSS WCHAN STAT TT TIME F COMMAND
0 24568 1 0 51 -20 84716 204 - U<s ?? 0:00.00 0 shutdown -r now
0 24514 1 0 31 0 84580 768 - U p1 0:00.00 5082 reboot

I did notice a samba process running collecting netbios info that I don't recollect starting:
0 24465 45 0 31 0 28080 996 - Us ?? 0:00.06 4000 /usr/bin/nmblookup -M -B 150.29.55.255

but it is probably innocent. I tried increasing yesterday (before the latest crash) the number of lookupd processes, but that is of no use as well. There is nothing in the system log regarding the lockup (except for the unlookupd process attempts to restart lookupd). Crashreporter I would think is not a suspect. Any suggestions as to what to try next?

Sorry, no suggestions, just a correction: I was referring to log entries added while CrashReporter had been disabled. There is no segmentation fault or bus errors when CrashReporter is running, only generic crash messages. (Not that either type help solving the problem at hand.)

This weekend I left a long calculation running after a fresh restart. I did not leave Mail or any web app running. I accessed the system from home using ssh. All worked fine for about a day and a half and suddently the machine was unreachable. This problem is infuriating. I suppose the next logical step is to disconnect the network cable, but perhaps someone else has a brighter idea?

Well so much for a shiny new G5. I now connect the ethernet cable for ten minutes at a time every few hours (as the problem always seem to occur after several hours of being connected). My (same) calculation is now three days running and I am actually getting some work done (the reason I have always bought macs to this point -- to get work done and not to have to screw around security/os problems). I just purchased a gigabit hub under the (superstitious) speculation that perhaps the problem was caused by either malformed packets or via automatic speed setting errors (but I have *no* evidence of them). In any case, I am now literally pulling the gigabit ethernet cable from the box most of the time which really sucks as I can't see the state of the calculations from home over the net (with the cable disconnected).
I did a more careful seach of the net, and sure enough on apple's mac os x server forumn there are a group of people with similar problems. The response from Apple seems to be that it *is* a lookupd problem, but they don't really seem to know what the cause is or how to fix it. At least one of the individuals there just order a linux box to replace the os x server, others are staying with 10.3. I am not so desperate (yet), I really like the mac os (when it works).
Has anyone heard any discussion of whether this is an issue being addressed with the 10.4.3 update? This is my hope from day to day (that and that the gigabit hub will fix the problem!)

Did the hub fix your problem?

I had nothing quite this bizarre happening, but there were idiotic anomalies no one could explain that all went away like magic when I got rid of an old router/switch combo and replaced with a fairly inexpensive new LinkSys router/switch.

Certainly the temporary fix of disconnecting the cable and not using ethernet (except for loopback) more than 15 minutes at a time works like a charm, but is hardly a permanent solution. It just tends to show that it is not a plist problem and really is a system level bug. The new gigabit ethernet hub is here and is working (when connected). There are two things I want to try when my current calculation is completed in a few days. First, I want to see if the new hub helps the situtation. I have (always) had a primary and seconary DNS server of my research institute. They both respond to ping quickly, I have always entered both (still working like a charm on this powerbook). I did read (somewhere) that having two specified might cause problems so I think I will leave only the primary. At this point, I am reduced to being almost superstitious and giving anything that sounds realistic a try. I already tried the unlookupd, I also made my own launchd script that flushed the cache of lookupd every hour. It didn't make any difference. The other possiblity is try diabling CrashReporter (which would really be unfortunate as I use its dumps to backtrack problems in the code I compile). If anyone hears anything about lookupd being "repaired" for the next update (or otherwise), it would give me hope!