From Bugzilla Helper: User-Agent: Mozilla/4.76 [en] (X11; U; Linux 2.4.1-ac20 i686) Up2date takes more than half an hour (I didn't wait for it to finish) while running on a 486 with 32MB of RAM over a cablemodem. Running with the same connection on a Celeron box with 128MB is fine. Reproducible: Always
what package set did you have chosen? Can you give me some memory statistics from top?
Seeing as I can't run up2date -u any more, and rhn_register is broken, I give up. [root@rushton /root]# rhn_register Traceback (innermost last): File "/usr/share/rhn/register/tui.py", line 1080, in ? main() File "/usr/share/rhn/register/tui.py", line 1076, in main tui.run() File "/usr/share/rhn/register/tui.py", line 1033, in run win = self.windows[index](self.screen, self) File "/usr/share/rhn/register/tui.py", line 627, in __init__ hardware.read_cpuinfo() File "/usr/share/rhn/register/hardware.py", line 106, in read_cpuinfo hwdict['speed'] = int(round(float(mhz_speed))) ValueError: empty string for float() [root@rushton /root]#
There is a config option for "headerCacheSize" that controls how many headers to cache in ram. The default is 40. Try lowering this to something like 10. Of course, this just trades swapping for lots of reading headers from disk, but on low mem machines, the later is probabaly faster.
FWIW, as another datapoint, I ran it on a 486 here; it upgraded vixie-cron, db3, popt, rpm, and rpm-python. Time was: real 5m9.777s user 4m5.990s sys 0m26.130s Certainly not fast, but not too bad. 486-50, 24MB memory, 42MB swap.
The issue about CPU speed on a 486 has been resolved and the fix is in CVS as well as the built package in the tree.