HighJinx: is doomviking feeling better?
up_the_irons i r not sure
if he's been down that long... that's.. not good
up_the_irons: they're the same as i7 i think
well with ecc ram support
i was surprised how fast my new i7 was :)
and e3s are often faster than e5s
unless you want to spend a lot of money.. e3 has higher clock rate than e5.. you need a very parallel work load for e5
that said... i5 versus i7 often shows similar speed in lots of things
did anyone else just find their freenode server went down?
actually now that i think about it freenode splits a lot
Any suggestions on how to fix my broken Ipv6 config?  http://pastie.org/5507258
mnathani: can you manually assign it?
mnathani: also, have you don the ifdown/ifup?
*done
I did restart networking
and rebood the box
how do you mean manually assign it?
ifconfig eth0 ...
don't remember the exact syntax
aren't those settings in my ifcfg-eth0 correct?
looked fine
Anyway, I'm heading out for a bit.. will try fixing it when I get back
mnathani: ifconfig eth0 inet6 add 2607:f2f8:af04::2/64
mnathani: Also I noticed there's no prefix length in the config...
but apparently that's unneeded
mercutio: interesting, so e3 is the way to go
mercutio: freenode splits all the time
does
FreeNode is a very large and popular target for DOS attacks, and when a node becomes sufficiently inundated and lagged, it disconnects from the network, hence a netsplit.
And now you know.
http://goo.gl/A4lPL
.--==*
love that ascii
.--==/*\
ha nice
up_the_irons: well it depends ...
up_the_irons: raw cpu isn't always the bottleneck on servers
up_the_irons: e5 also has some low latency stuff for i/o interconnects, and supports more ram, and quad channel memory
like if you want to go above 32gb of ram...
http://www.intel.com/content/www/us/en/io/direct-data-i-o.html
also if you pay "per cpu" a lot of money for licensing you can get really fast really expensive e5s
mercutio: yeah makes sense
lots of people want to save money, but if you get e5-2620*2 on one server versus two e3-1270v2 servers, it's probably similar cost
e5 has 12 cores with hyperthreading, two 3s have 8 cores with hyperthreading
but e3 is like 3.4ghz and e5 is like 2.1 ghz
hmm 2 ghz burst to 2.5
versus 3.5 ghz burst to 3.9 ghz
but if you're not doing highly parallel applications, single thread performance is better on e3
oh btw up_the_irons i noticed there are less bgp routes now, were you hitting your limit again?
mercutio: when Any2 had problems a couple weeks ago, probably some peers dropped off (and haven't reset their session)
ahh
cos they don't want to risk instability :)
yep
we got the RFO for that.  It was a multicast flood.  Supposedly, _now_ they inserted a rate limit and ACL for multicast.
joy
the thing is they have lots of peering exchanges!
so they don't relaly have an excuse
sometimes i'm surprised the internet works as well as it does
s/some/many/
I find the Internet works about the same as driving - It's mostly sheer luck and self-preservation.
(that there aren't more accidents)
brycec: heh
i think it's funny when china manages to somehow take a whole lot of US routes
"somehow"
That's not the only case really.
a lot of problems come from least cost routing
one year at defcon the network peeople routed everything out through new york via bgp (from vegas)
there's no in built distance / best path stuff
there's just a list of whose network it goes through
so if you advertise someone elses ip addresses as directly connected then that can take precedence over someone else who's learned through someone else
yep
http://techtalkin.com/Thread-Modifying-cPanel-s-default-zone-template-to-add-correct-SPF-records very nice
I love getting mailer-daemon bounces "your message could not be delivered because it failed SPF validation. Fix your SPF!"
They're all worded like it's my SPF that's broken, when it's the exact opposite.
can't stay away from you guys lol
u love us dr_jkl
up_the_irons: you around actually? i have a question
if i can pm.
:)
I ran into a weird problem on a freshly-reimaged fbsd 9.0 VPS. 'portsnap fetch' failed, complaining that it could not create the relevant files in /var/db/portsnap/snap. After the failure, there were exactly 11999 files in that directory. I could create a new file in it with touch, but (for example) vi would fail to start because it couldn't create temp files. I rm -rf'ed /var/db/portsnap/* and /usr/ports, and then portsnap fetch
completed without issue
has anyone seen something like that before? I forgot to df -i before I deleted those directories
CESSMASTER: how big is your disk?
5G
df reported it was 50% full
ah yes
you ran out of inodes most likely
and 'touch' doesn't actually use an inode if the file remains empty?
hmm
i mean, the fact that it crapped out with 11999 files in that directory makes it look like some kind of weird per-dir limit, except i was trying to run vi on /etc/portsnap.conf so presumably the temp file would be created in a totally different dir
what static said
there have been issues in the past with freebsd default inodes 5gb and ports
i guess the UFS guys don't care about such small partitions too much
CESSMASTER: this guy has a good explanation of the problem:
http://gromnitsky.blogspot.com/2011/08/freebsd-9-newfs-block-size-and-fragment.html
heh
up_the_irons: got a sec?
dr_jkl: don't ask to ask, just ask (and if I'm busy, i won't answer until i have more time)
dr_jkl: or i'll refer you to support :)
heh sure i just didnt want to be improper about just randomly messaging you
dr_jkl: it's just irc and i'm not an emperor ;)