#arpnetworks 2015-05-06,Wed

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)

WhoWhatWhen
***m0unds__ has quit IRC (Ping timeout: 256 seconds)
m0unds_ has joined #arpnetworks
[00:11]
................. (idle for 1h22mn)
LT has joined #arpnetworks [01:35]
..... (idle for 24mn)
b^_^d has quit IRC (*.net *.split)
dwarren has quit IRC (*.net *.split)
b^_^d has joined #arpnetworks
dwarren has joined #arpnetworks
[01:59]
..................................................... (idle for 4h21mn)
plett has quit IRC (Ping timeout: 276 seconds) [06:22]
.... (idle for 18mn)
plett has joined #arpnetworks [06:40]
toeshred has quit IRC (Read error: Connection reset by peer) [06:47]
brycecBut not for long, 5.8 has disklabel templating :D
of course that only works on the initial install...
[06:54]
..... (idle for 23mn)
mkbon BSD you have to resize the old fashioned way: dump, newfs, restore [07:18]
***toeshred has joined #arpnetworks [07:32]
........................... (idle for 2h12mn)
LT has quit IRC (Quit: Leaving) [09:44]
................ (idle for 1h15mn)
RandalSchwartznot freebsd zfs
it'll recognize larger disk
[10:59]
........... (idle for 50mn)
***b^_^d has quit IRC (*.net *.split)
dwarren has quit IRC (*.net *.split)
hazardous has quit IRC (*.net *.split)
NiTeMaRe has quit IRC (*.net *.split)
meingtsla has quit IRC (*.net *.split)
hazardous has joined #arpnetworks
meingtsla has joined #arpnetworks
b^_^d has joined #arpnetworks
dwarren has joined #arpnetworks
NiTeMaRe has joined #arpnetworks
[11:49]
...... (idle for 28mn)
hazardous has quit IRC (*.net *.split)
hazardous has joined #arpnetworks
[12:20]
mercutiobrycec: disklabel templating/
oh to say how much to use of partitions etc
[12:27]
............... (idle for 1h13mn)
brycecyeah [13:41]
..... (idle for 21mn)
mercutioyeah i saw discussions about the autoinstall stuff on the mailing list. [14:02]
i wonder if some lvm type solution is going to come to openbsd [14:11]
mike-burnsIs that different from bioctl?
(I don't actually know.)
[14:12]
mercutiolvm is about sub volumes
that you can resize at runtime
[14:12]
mike-burnsAh. [14:13]
mercutioso you can allocate various amounts of space, but change them later.
bioctl is about managing raid systems
lvm is the kind of layer that it seems to make most sense to implement "cache" devices too
[14:13]
mike-burnsHeh, the misc@ thread where someone asked about lvm for OBSD mentions Arpnetworks.
http://marc.info/?l=openbsd-misc&m=132100433813215&w=2
[14:14]
BryceBotopenbsd-misc: "Re: similar lvm tool on openbsd??" from carlopmart <carlopmart () gmail ! com> @ 2011-11-11 9:35:23 [14:14]
mercutioheh [14:15]
plettAren't zfs volumes roughly equivalent to LVM LVs? [14:16]
mercutioplett: not really
you can fix the same issue
but zfs is kind of a much bigger more complicated solution
i'm not a huge fan of lvm, but it is convenient
[14:16]
mike-burnssoftraid(4)? [14:17]
mercutiosoftraid last i knew was just joing 2 or more partitions to form a pretend partition [14:17]
plettsoftraid would be md raid in Linux [14:17]
mercutioit doesn't do chunking
some raid systems do support doing it there though
[14:17]
RandalSchwartzanyone taken freebsd 8 to 9?
I'm reading the release notes... doesn't look too crazy.
[14:23]
the 9.3 install notes make it appear I can go direct from 8.4 to 9.3
I'll have to rewrite some of my rc.conf for networking eventually though
maybe even soon
[14:36]
brycec(IIRC softraid also incorporates some cryptsetup-type stuff, comparing OpenBSD and Linux-isms)
RandalSchwartz: I did... years ago :p
[14:37]
....... (idle for 30mn)
***himuraken has quit IRC (Read error: Connection reset by peer) [15:07]
RandalSchwartzwell, I'm practicing with a scratch box before doing a real box. :)
going to 9.3 first, since the EOL of both 9.3 and 10.1 are the same.
[15:09]
.......... (idle for 48mn)
wow ... "Preparing to download files..." is taking forever [15:58]
ok.. movement now
"fetching 8763 files"... hehe
[16:07]
..... (idle for 24mn)
freebsd-update install... no turning back now. [16:31]
kernel updated, now updating userland [16:43]
mercutioyeh it's pretty slow [16:44]
RandalSchwartzwell - changing a lot of files [16:44]
mercutioare you going to upgrade to 10 afterwards? [16:44]
RandalSchwartzno point just yet [16:44]
mercutiook [16:44]
RandalSchwartzsince EOL of 9 and 10 are the same
I try to take minimal risk steps
I have 6 systems I have to do this to. :)
[16:44]
mercutioyip
i feel the same way about important systems
[16:45]
RandalSchwartzif only I was a linux lover, I could use CoreOS for all this :) [16:45]
mercutiobut non-important i like to be bleeding edge :)
i haven't checked out coreos yet, someone else was mentioning that i should look at it
[16:45]
RandalSchwartzmaybe me. :) [16:46]
mercutionah it was someone i know in real life :)
i love it how you go to coreos.com and it doesn't tell you what it does :)
[16:46]
RandalSchwartzmy strategy for this was: install 8.4 on scratch box. build 8.4 packages from our current origins and options. upgrade to 9.3 (happening now). build 9.3 packages. [16:47]
mercutioahh overview [16:47]
RandalSchwartzonce all that is done, take a live box, upgrade 8.4 to 9.3, and install my 9.3 packages.
if that works, proceed 5 more times. :)
[16:47]
mercutioyeh it's much messier than it should be.
you can't just upgrade ready to go to a new zfs mount point
check it's all working fine then swap over to it
then roll back if it doesn't work out leaving the old system in place
[16:47]
RandalSchwartzyeah, I'm planning on installing beadm someday soon
that would effectively do that
[16:48]
mercutiocool [16:49]
RandalSchwartz... http://www.freshports.org/sysutils/beadm/
... http://www.linuxquestions.org/questions/%2Absd-17/howto-zfs-madness-beadm-on-freebsd-4175412036/
[16:50]
mercutiouhh this is just like what i was saying ? :) [16:52]
.... (idle for 18mn)
RandalSchwartz9.3 up and running! now building the 9.3 poudriere
so glad I found out and started using poudriere
building individual ports on each machine seemed nuts :)
696 packages to build
[17:10]
mnathani_is this true? : curl is 17 years old and maintained by one person [17:19]
mercutiomuch outside help [17:19]
RandalSchwartzsounds likely
I interviewed him for floss
[17:19]
mercutiohe's only recently being paid to work on it too
i mean being able to work on it at his day job
i think much humbleness too
[17:19]
mnathani_found it http://twit.tv/show/floss-weekly/51 [17:20]
RandalSchwartzas I recall, he reminded me of richard hipp
sqlite and fossil
[17:20]
mercutio17 years ago is about when there was heaps of new projects that went somewhere
randal: as in the dos fossil drivers?
[17:20]
RandalSchwartzno... fossil the SCM [17:21]
mercutiooh [17:21]
RandalSchwartz... http://twit.tv/show/floss-weekly/320 [17:21]
mercutiomnathani_: postfix was written/maintained by one guy for nearly 17 years too.
again some outside help
open source is good like that :)
but the postfix author always impressed me with his dedication
if a user is confused by something, he blames the system
as being too complicated
[17:21]
RandalSchwartzthe ntp code is maintained by one guy
postfix is wietse, right?
[17:23]
mercutioa lot of things are easier with one guy
randal: yeh
[17:23]
RandalSchwartzsame guy who did some security tester [17:24]
mercutiohe seems impressive
but now arrogant
[17:24]
RandalSchwartzmet him some years ago
oh... right... he did crack
[17:24]
mercutiowhat? [17:24]
mnathani_RandalSchwartz: are you actively using vagrant? [17:24]
RandalSchwartzmy name is in the crack manual [17:24]
mercutiooh not the drug [17:24]
RandalSchwartzthe password brute force [17:24]
mercutioheh [17:24]
RandalSchwartzat the very moment, no.
but earlier today yes
$client requires it for dev env
[17:25]
mercutiowhat [17:25]
RandalSchwartzhope to move to docker soon [17:25]
mnathani_what hypervisor are you using it with? [17:25]
mercutiowitese moved from ibm to google
does that mean postfix is going to the dark side?
[17:25]
RandalSchwartzvirtualbox [17:25]
mercutio(when did google become more evil than ibm?) [17:25]
mnathani_osx hosts? or linux [17:25]
RandalSchwartzosx [17:25]
mnathani_I recently started using it with windows host and virtualbox [17:26]
RandalSchwartzworks on linux now? [17:26]
mnathani_vagrantup.com has binaries for deb and rpm so probably [17:26]
RandalSchwartzwow
times, they a changin'
[17:26]
mnathani_have you considered bringing in the developer: Mitchell Hashimoto for a floss show? [17:27]
RandalSchwartzwho is that?
and what project?
[17:28]
mnathani_the vagrant author [17:28]
mercutioi assume it's vagrant
because he was jusut talking about vagrant :)
[17:28]
RandalSchwartzoh... that'd be cool.
email him, tell him to email me.
[17:28]
mnathani_will do [17:29]
RandalSchwartz... Since version 1.6, Vagrant natively supports Docker containers, which serve as a substitute for a fully virtualized operating system.
ooooh
oooh ooh oh
too bad that's just for linux
why isn't there a freebsd docker?
jails were the inspiration for LXC
[17:29]
mercutioi thought openvz would be the inspiriation
and linux-vserver
[17:30]
RandalSchwartz.. http://unix.stackexchange.com/questions/127001/linux-lxc-vs-freebsd-jail [17:31]
mercutiowikipedia never seems ot make it easy to find out when projects start :(
linux-vserver oct 2001
[17:32]
RandalSchwartzjails started in freebsd 4.0 in 2001
2000
[17:33]
mercutiohmm [17:33]
RandalSchwartzdarn it.. you typed 2001 just as I was trying to type 2000, and my brain rewrote it [17:33]
mercutioheh
damn whenever i try and track when things started i feel older :/
[17:33]
RandalSchwartzhere's a history: http://www.cybera.ca/news-and-events/tech-radar/contain-your-enthusiasm-part-two-jails-zones-openvz-and-lxc/ [17:33]
mercutiochroot is older of course [17:34]
RandalSchwartzsure...
the other article talks about that.
[17:34]
***himuraken has joined #arpnetworks [17:40]
.......... (idle for 48mn)
NiTeMaRemercutio: for awhile now maybe a year?
dec/2014
[18:28]
mercutiodamn what did i ask? [18:29]
NiTeMaReaugust 2014 actually
solusvm ipv6 subnet
[18:29]
mercutiooh dec 2014 is pretty late.
even 2014 is pretty late
[18:30]
NiTeMaReya
took them awhile
[18:30]
mercutiofrom what i've seen of solusvm ipv6 seems like bad idea [18:30]
NiTeMaReit works pretty well now, even got ebtables working nicely with it
i only use KVM with it not xen/whatever else though so i'm not sure about other virts
[18:31]
mercutiothey tend to spam the network with heaps of nd/arp
it's ironic that arp doesn't have arp issues :)
[18:32]
NiTeMaRelol [18:33]
mercutiolike choose a random vm someewhere cheap on lowendbox say
then do tcpdump -p -l -n -i eth0 ! port 22 on an idle server
and there'll be like 250arp/sec etc often
and lots of other crap
if you have 20 servers, that's 5000arp/sec...
[18:33]
NiTeMaReya [18:34]
mercutiolinux 4.1 is bringing a new cpu thing to linux for kvm where it can bsaically stop the kernel running while it's ruunning a vm
but even if arp's are reasonably cheap to process they "wake up" the cpu a lot
and thrash cpu caches etc.
it's just one of many virtualisation issues though
[18:34]
NiTeMaRei should read up about 4.1
haven't done it yet
[18:35]
mercutioand it seems likely that intel cpus will just make context switches cheaper or divide cache better or such
i wouldn't worry too much yet.
there's nothing earth-shattering.
like the kvm optimisation is nice, but it's not going to make a /huge/ difference
it's more in getting closer and closer to 0 overhead for cpu bound tasks
and where kvm seems to struggle more is disk/network
it's like upgrading the cpu on a disk bound database server.
it may make some fast things faster, but it won't stop "slowness"
[18:36]
NiTeMaRebottlenecking :P [18:39]
mercutioyeah once people start doing 10 gigabit more they'll notice kvm's network issues more.
it's kind of cpu hungry at gigabit, and struggles a bit with high udp loads.
[18:39]
RandalSchwartzsteve gibson calls that the background radiation of the net [18:40]
NiTeMaRe- EXT4 now supports file-system level encryption after being a feature driven by Google for Android.
that will be nice
[18:40]
mercutionite: are you really going to trust ext4 encryption on initial realease?
release
[18:40]
NiTeMaReno
it will be nice once it is developed though
[18:41]
mercutioi still want lz4 compression on ext4, xfs etc.
i've started shifting from ext4 to xfs.
[18:41]
NiTeMaRei've been wanting to play around with btrfs [18:41]
mercutioi tried lzo compression with btrfs, but btrfs is just too unstable, even for non-crtical systems. [18:41]
NiTeMaReit is still unstable? [18:42]
mercutioyeah [18:42]
NiTeMaRei saw opensuse implemented it native [18:42]
mercutiowell it depends, it's unstable using it as /
on a ssd with a small partition
well it was ssd raid, but same diff
opensuse implemented reiserfs native too
[18:42]
NiTeMaReinteresting [18:42]
mercutioi used to use reiserfs for squid
but i never trusted it with critical data.
[18:43]
NiTeMaReyou've played around with bhyve much? [18:43]
mercutionah not yet
freebsd is too unstable for me to dare :)
[18:43]
NiTeMaRei love freebsd [18:44]
RandalSchwartzI'm presuming once I get my ARP box to 10.x I can use bhyve
my DigitalOcean 10.x box can't do it
[18:44]
mercutiofreebsd is ok as long as you tread lightly
but i'm just too paranoid about a system that can't deal with 9k mtu's without running out of memory
[18:44]
RandalSchwartzZFS on / for the win [18:44]
mercutiobasically their memory allocator gets fragmented and spaz's
it's ok if you use 2k mtu
there's some other weird stuff too
[18:45]
NiTeMaReI fly out to Texas Sunday and I still have 0 packed
should probly get on that
[18:51]
mercutioheh i never pack in advance
i always think i /should/ pack in advance
[18:56]
RandalSchwartzif you pack afterward, it's not as effective.
lots of loose things on the security belt
[19:01]
m0undslol [19:13]
.... (idle for 15mn)
RandalSchwartzyou drag the six plastic tubs with you to the plane. :) [19:28]
m0undsi just bring a dolly and my dresser [19:31]
RandalSchwartza garry dolly? [19:33]
m0undshaha
i don't think up_the_irons would want to carry my dresser around
[19:33]
RandalSchwartzheh [19:35]

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)