***: vtoms has joined #arpnetworks
heavysixer has joined #arpnetworks
visinin has quit IRC ("nyc") cablehead: up_the_irons: i spoke to joel about it, he said unless there was something in it for slide, it was unlikely they'd peer
up_the_irons: pretty lame :(
up_the_irons: he mentioned something about a communal hub everyone can just plug into? mhoran: up_the_irons: Are you in on the ANy2 exchange?
Ah, yes you are. :)
So did Coresite rename Wilshire Annex to 900 N. Alameda when they rebranded? ***: ballen|away is now known as ballen
ballen is now known as ballen|away
ballen|away is now known as ballen up_the_irons: cablehead: lame :( the communal hub he talked about is probably the Any2Easy exchange, a way to make peering automatic, but is not always ideal b/c you pick up everyone, or no one. Think auto-config vs. manual config. some people (including myself) are not always comfortable w/ peering indescriminatenly like that
cablehead: I'm part of Any2Easy as well, but I prefer direct peer if possible. but if he won't do it, oh well
mhoran: yeah, i'm on Any2 :) cablehead: up_the_irons: yeah really lame :( up_the_irons: mhoran: i believe 900 n. alameda is still known as wilshire annex mhoran: up_the_irons: I just can't find any reference to it on their site anymore. But searching for the annex brings up the page. up_the_irons: funny ***: ballen_ has joined #arpnetworks
ballen has quit IRC (Read error: 148 (No route to host)) up_the_irons: mhoran: they probably want everything to be known as "coresite" now, and the different locations are just different locations (they don't have individual names anymore). that's what i would think mhoran: Yeah. up_the_irons: Network Next Hop Metric LocPrf Weight Path
* 208.76.68.0/22 206.223.143.103 0 189 0 13949 i
*> 206.223.143.103 0 190 0 13949 i
* 208.87.172.0/22 206.223.143.103 0 189 0 13949 i
*> 206.223.143.103 0 190 0 13949 i
s1.lax#
cablehead: ^^ looks like I get Slide's routes through Any2Easy anyway. not ideal, but better 'n nothing :)
cablehead: ping slide.com from your VPS, you'll probably be quite impressed with the ping time ;) ***: vtoms has quit IRC ("Leaving.")
vtoms has joined #arpnetworks
ballen_ has quit IRC (Read error: 113 (No route to host))
Thorgrimr has joined #arpnetworks
heavysixer has quit IRC ()
heavysixer has joined #arpnetworks
vtoms has left
heavysixer has quit IRC ()
vtoms has joined #arpnetworks Thorgrimr: date up_the_irons: fail Thorgrimr: Damn, no takers ***: mxmasster has joined #arpnetworks up_the_irons: new blood!
mxmasster: welcome to #arpnetworks mxmasster: hey Gary
reading your twitter up_the_irons: haha
cool mxmasster: figured I'd join and say hi up_the_irons: hi :) mxmasster: hear back from nasa? up_the_irons: no :( ***: baklava has quit IRC (Read error: 54 (Connection reset by peer))
heavysixer has joined #arpnetworks
mxmasster has quit IRC ()
vtoms has left
vtoms has joined #arpnetworks
bobbyw has joined #arpnetworks bobbyw: hey guys ***: vtoms has quit IRC ("Leaving.") bobbyw: anyone active here? ***: heavysixer has quit IRC ()
baklava has joined #arpnetworks
ballen has joined #arpnetworks ballen: what up jeev: put the money in the bag ballen: what bag in where now jeev: put your monies in my bag ballen: [FBI]: off ***: [FBI] has left
[FBI] starts logging #arpnetworks at Thu Sep 17 21:23:15 2009
[FBI] has joined #arpnetworks jeev: not really ballen: then don't download it
see if I care jeev: i'll exploit that mp3
and put rap on it ballen: you're going to exploit a rapper ? jeev: no
i'll haxor that mp3
and overwrite it with gunit and mozart remix ballen: ;-)
quite a bit of rap in it actually, just remixed mike-burns: Is Gunit a GNU unit testing tool?
I know that Mozart is an implementation of the Oz programming language. ballen: or a gnu wanker jeev: http://www.theregister.co.uk/2009/09/15/killer_robot_killed_by_fighter_jet/ ballen: 14 hellfires, bitchin mike-burns: That's quite close to Dr. Strangelove. ballen: someday... someday mike-burns: It's a step in the right direction. Or wrong, or whatever. ballen: reminds me of some cheesy sci-fi flick
this killer-robot fucked some serious shit up though ***: bobbyw has quit IRC ("Leaving.")
bobbyw has joined #arpnetworks bobbyw: is there an arp rep here? jeev: he should be by soon
anything we could help you with ? bobbyw: eh, thinking about switching from linode jeev: i've only heard about em, never tried ballen: I switched from Slicehost and haven't looked back bobbyw: actually slicehost, yeah I've used about every vps service ballen: but then again I'm very pro FreeBSD and was gritting my teeth using Linux bobbyw: yeah, I think deep down I prefer freebsd
philosophically jeev: yea i prefer bsd too but
not on a vps bobbyw: why not? ballen: everyone deep down prefers FreeBSD
they just don't know it bobbyw: lol
yeah jeev: on vps? eh
i dunno, building world
just think it'd be too slow ballen: who really needs to build world
these days at least jeev: I DO! ballen: no you don't
why
do you
need
to
build WORLD bobbyw: linux is really a hot mess from all angles jeev: cause
I'M
DEAD
SEXY
bobbyw, need any info about arp? ballen: that you are, anywho bobbyw: jeev, what is the uptime guarantee? ballen: thats a good question for up_the_irons bobbyw: ok
I'll come back another time ballen: its a small operation bobbyw: I have about a month
ballen, that's what worries me
small ops usually means more downtime ballen: yea... depends on how critical your VPS's are
and how much you want to pay for uptime, vs value of ARP bobbyw: yeah, definitely a trade-off, I am just curious more than anything
not super anal about it ballen: right
I know he's working on a DRBD solution bobbyw: drbd? ballen: which mirrors VMs to another box bobbyw: ah jeev: hmm good question ballen: distributed ... block device jeev: i dunno ballen: so basically if there was a need to take down a box the downtime would be the time of a restart jeev: only downtime ive had this month was the planned upgrade
45 minutes at night bobbyw: yeah that's cool, and I am used to that ballen: yea Garry has a 41 minute hw upgrade this month
had*
and I think one requires restart for kernel upgraded
required*
that might of been last month
though
the kernel that is bobbyw: thanks for the info guys, I'm out for now, I will come back another time to chat with one of the arp dudes ballen: mmk jeev: try back in an hour
;D ballen: yea he's usually around late bobbyw: ah ok
later ***: bobbyw has left ballen: peace jeev: i just showered
why do my nadz smell ballen: jesus jeev: my friends box got hax0red today
cause he was running comusthumbs.com
php hax0r
everything has <script> viriI in it ballen: nice
my brain is freaking mush right now
so freaking annoying
so much writting of the thesis jeev: i need to rest ballen: likewise up_the_irons: omg, u guys been chattin' -: up_the_irons reads scrollback up_the_irons: ballen: fyi, the DRBD solution, once finalized, would allow me to take down a box with *zero* VPS downtime; the VMs would be live migrated to another box. I've got all this working except for the "live" part ;) for some reason the emulation doesn't take on the target host. So right now, it requires a VM halt / boot (still a lot less downtime) ballen: hows the live migration part actually work
as far as memory... network connection hand off
etc so forth up_the_irons: ballen: memory image is transferred; while the transfer is happening, dirty pages are marked (pages written to since the start of transfer); after transfer, emulation is halted and dirty pages are transferred (very quickly b/c there are a lot less of 'em), emulation is restarted on target. An ARP broadcast is sent to tell switches / routers of the new location of the IPs bounds to the VM
interfaces
that's about the jist of it ballen: curious, do your two test servers that you're trying to live migrate have identical hardware? up_the_irons: ballen: almost ballen: same CPUs? up_the_irons: ballen: yeah ballen: hmm
so is live migration something out in a wild with KVM, or you're own doing up_the_irons: ballen: http://www.linux-kvm.org/page/Migration
ballen: all KVM/QEMU ballen: neato up_the_irons: supposedly this support has been around for a while and people have reported a lot of success
so i'm surprised it is not working for me ballen: whats it currentl failing on
btw having live migration will kick the shit out of Slicehost, et. al up_the_irons: ballen: yeah i know, which is why i want it :)
ballen: on the target, after the migration has finished, I get tons of this in syslog:
emulation failed (pagetable) rip ffffffff807b6dbe ff ff 08 00 ballen: do you by chance have another box to test as a target obsidieth: can someone
download and dcc me a file up_the_irons: not really, that's the thing... obsidieth: my internet is capped and it keeps stalling jeev: heh ballen: does this file happen to be located at some internal NASA website? obsidieth: its on some russia rapidshare
its a gnome theme that i really want :( up_the_irons: i may just need to update my kvm / qemu packages, but that's something I don't take lightly
given the current kvm / qemu have been running really damn well :)
NASA never got back to me jeev: vps's are the future eh up_the_irons: <sniff> obsidieth: http://www.sharemania.ru/0133756 jeev: i dont think future for single customers
i think future for corps and businesses, even present i guess ballen: up_the_irons: yea I was gonna say may want to try upgrading up_the_irons: ballen: yeah, that's really the next logical step ballen: http://sourceforge.net/tracker/index.php?func=detail&aid=2778112&group_id=180599&atid=893831
not much of resolution, but similar errors
also: http://sourceforge.net/tracker/index.php?func=detail&aid=1852938&group_id=180599&atid=893831
here we go: http://sourceforge.net/tracker/index.php?func=detail&aid=2804562&group_id=180599&atid=893831 up_the_irons: the 2nd one is quite old
oh ballen: your KVM versions newer than that bug fix?
July 01 up_the_irons: i don't get "load of migration failed" though
i'm on qemu 0.10 and kvm-84 ballen: hmm
I assume both source and target are 64 bit systems? up_the_irons: yep ballen: not trying to install Windows 98 are you?
;-) jeev: up_the_irons, you had a possible client in here a bit ago up_the_irons: ballen: hahahah, no
jeev: yeah, I saw in the scrollbabck
he'll be back it seems
i should try the migration with: -no-kvm-irqchip or -no-kvm-pit ballen: https://bugzilla.redhat.com/show_bug.cgi?id=508801
what kernel are you using up_the_irons: 2.6.28-15
everything just keeps pointing to "this was fixed in this commit", which is newer than what I have ;) ballen: heh
http://patchwork.kernel.org/patch/18599/ up_the_irons: i need to build me another test box ballen: also check commit: 323d3b06db8bf2d8e4c5ed1a390668ae7b1b84bf
heres the diff for that commit http://pastie.org/621423
was mentioned somewhere
looks like it touches "Cannot emulate" error message
this all looks related to the patch: http://osdir.com/ml/emulators.kvm.scm/2007-09/msg00135.html
so moral of the story.... upgrade up_the_irons: yup, upgrade :) ballen: curious are you having the VM directly use the DRBD device or a file on the DRBD for a disk up_the_irons: ballen: DRBD volume is defined that the VM uses directly; the backing store for the DRBD volume is an LVM volume ballen: cool
figured up_the_irons: basically DRBD keeps /dev/vol1/foo in sync on hostA and hostB ballen: yea I know
just seeing how you were doing it
http://www.convirture.com/products.html
interesting product
free up_the_irons: i really shy away from products that force the GUI
GUIs can't be scripted, and I don't wanna monkey click all day
;) ballen: yea just looking at it for VMware center replacement type ideas
VCenter* up_the_irons: i c
I use libvirt
(command line), but there are GUI tools built on top of it if you need that
like, virt-manager, i believe ballen: yea
in your test bed, Intel or AMD cpus? up_the_irons: Intel ballen: ah damn
http://patchwork.kernel.org/patch/43611/
AMD patch -: up_the_irons is building a FreeBSD kernel up_the_irons: i c
anyone know how long it takes to build a GENERIC kernel, typically? ballen: 10-15?
minutes
depending on hardware
http://feanor.sssup.it/~fabio/freebsd/lkvm/
anyone actually try KVM running on FreeBSD up_the_irons: not at all ballen: bleh wish more effort would go into virt using BSD up_the_irons: yeah, it's lacking