April 18, 2001 - University of North Carolina reported
"our performance degraded continually
until we came again to a complete halt this morning at 7ish EDT. after
a reboot and an investigation, we're back up and up much more quickly
thanks to the changes jonathan made yesterday. we're still looking
at the problem which seems to be load related. and we have moved red
hat 7.1 to a different server." Well, that's why you couldn't
get to our site. I could understand that message, but then this arrived:
"we were running red hat 7 with a VA 2.2.18 kernel that allows
ext3 file
systems on divahouse the www.ibiblio.org machine which is also the
major
disk farm for the complex. all this on an ibm netfinity with mucho
memory.
this weekend when a very desirable distribution was released (red
hat
7.1), we saw our outbound traffic get very high (normal) but we
also saw
the server crawling to a halt followed by a lockup.
the wonderful journaling for ext3 was not working right so late
yesterday
after reinstalling the root partition completely and reinstalling
the
kernel, jem also turned off ext3 and went back to ext2.
last night the outbound traffic climbed again and the slowdown
began
again. by 7 edt this morning we took a kernel panic. a reboot kept
us up
for about another hour and a half. then we got another kernel panic.
dls has nuked the red hat 7.1 release. he is now recapturing it
on jungle
so it will be on its own server and so not hurting others. i also
asked
hte campus networking folks to look at enabling a governor on the
machine
so's we's be back down at about 75Mbs.
jem and jreuning are looking at kernel alternatives at the moment
(red hat
2.2.19 and va 2.2.18.va2.1).
more news as it happens. right now we look pretty good (fingers
crossed
for good luck)"
Did you get all that? I didn't. And then: "the unc campus network
was down for a couple of hours today between noon
and 2:30 edt (us)."