***: gift_ has quit IRC (Ping timeout: 276 seconds)
gift_ has joined #arpnetworks
Ehtyar has joined #arpnetworks ant: up_the_irons: i remember having heard about using dhcpv6 prefix delegation to assign prefixes to cpe. but never seen this in reality as ipv6 is currently nor really available for normal home customers in germany
oh, yeah. some isps are running tests... ***: heavysixer has quit IRC (Remote host closed the connection)
heavysixer has joined #arpnetworks
ChanServ sets mode: +o heavysixer
gift_ has quit IRC (Quit: it puts the packet in the socket or else it gets the close() again)
Hien has quit IRC (Read error: Connection reset by peer)
Hien has joined #arpnetworks
mike-burns has quit IRC (Read error: Connection reset by peer)
mike-burns has joined #arpnetworks
ChanServ sets mode: +o mike-burns
mike-burns has quit IRC (Read error: Connection reset by peer)
mike-burns has joined #arpnetworks
ChanServ sets mode: +o mike-burns
heavysixer has quit IRC (Remote host closed the connection)
heidar has quit IRC (Quit: Quit)
heidar has joined #arpnetworks
heavysixer has joined #arpnetworks
ChanServ sets mode: +o heavysixer
CaZe has quit IRC (Remote host closed the connection)
CaZe` has joined #arpnetworks
CaZe` is now known as CaZe
CaZe has quit IRC (Remote host closed the connection)
mike-burns has quit IRC (Read error: Connection reset by peer)
mike-burns has joined #arpnetworks
ChanServ sets mode: +o mike-burns
CaZe` has joined #arpnetworks
CaZe` is now known as CaZe
heavysixer has quit IRC (Remote host closed the connection)
ChurchyLaFemme is now known as jdoe
HighJinx has quit IRC (Quit: Computer has gone to sleep.)
HighJinx has joined #arpnetworks RandalSchwartz: are we able to choose our own iso's in the cdtray, or does that still take a support request?
I'm uplifting my machines to 8.3, and want to be sure the DVDs are in sync
especially if I upgrade my zfs versions
hmm. I don't see any menu item for that
only reverse DNS
ok, support request time. :) mikeputnam: someone needs to patch mount(8) to accept the device from a pipe! curl -f ftp.site.com/foo.iso | mount /cdrom
;)
actually: curl -f ftp.site.com/foo.iso | mount - /cdrom jdoe: still have a machine stuck on 8.1
not sure if I want to go to 9 or 8.3... ***: mike-burns has quit IRC (Read error: Connection reset by peer)
mike-burns has joined #arpnetworks
ChanServ sets mode: +o mike-burns -: RandalSchwartz sends email to support plett: ant: We (AAISP, a .uk ISP) use DHCPv6 and DHCPv6-PD in production to get IPv6 to DSL customers, so it is in use ***: heavysixer has joined #arpnetworks
ChanServ sets mode: +o heavysixer
urkle has joined #arpnetworks urkle: heya, up_the_irons you around? ***: RandalSchwartz has quit IRC (Ping timeout: 245 seconds)
RandalSchwartz has joined #arpnetworks RandalSchwartz: wow... my box slowed to a crawl starting 20 minutes ago
anything going on with kvr05?
I know I'm beating up kvr24 with a freebsd uplift ***: RandalSchwartz has quit IRC (Ping timeout: 260 seconds)
RandalSchwartz has joined #arpnetworks
RandalSchwartz has quit IRC (Changing host)
RandalSchwartz has joined #arpnetworks RandalSchwartz: wow... all of a sudden my disk is back to normal
that's weird
I was gtting 50k/second disk writes. :(
now I'm getting 400MB/sec jdoe: that happens to me semi-often.
... the former, not the latter ;) up_the_irons: RandalSchwartz: for stonehenge vps? yeah, i had to kick a vps gobbling disk I/O on kvr05 -: jdoe wonders if IO performance is likely to be better on a node with more large or small instances. jdoe: ... assuming up_the_irons doesn't distribute them, that is. up_the_irons: jdoe: i have found that there is no correlation. a large vps can do a lot of disk I/O, or many small ones can do the same
it really depends on the users
if, by some randomness, a host has mostly, say, dns, vpn irc, gamer, customers, then disk I/O is pretty low tooth: <--
irc and email, tiny web.
:-) RandalSchwartz: ahh!
so it was unusual
and you did take action
sounds like I ought to add a zabbix alert based on speed of disk I/O :) up_the_irons: RandalSchwartz: yeah, i have a load monitor; got the page _right_ as i left the data center (for another thing; replaced failed disk in kvr02), and then was on the road for an hour... a lot of times these things cure themselves, but sometimes not -: up_the_irons loves tiny customers like tooth ;) jdoe: haha.
can't remember if I can claim tooth or if he found this somewhere else. tooth: wat
no, i found this on my own. -: jdoe claims tooth anyway up_the_irons: LOL jpalmer: my vps is mostly just an IRC service now. tmux+irssi. I don't even run DNS on it anymore. hehe RandalSchwartz: wow. two hours to makeworld on freebsd 8.3
and I have 4 more VPS to go
make buildworld that is jpalmer: RandalSchwartz: can you export the buildworld, and just installworld at the others? RandalSchwartz: not sure jpalmer: well, you can. I mean, is it an option :P RandalSchwartz: I don't see why not... would I just rsync /usr/src from one machine to the next? jpalmer: I usually NFS share /usr/src and /usr/obj so rsyncing those two should work as well.
assuming same arch, and such. RandalSchwartz: yeah, these are all arp VPS
I presume they're pretty much identical jpalmer: I used to buildworld on a quad core machine, and installworld to hundred of Alix and WRAP machines. RandalSchwartz: freebsd? jpalmer: yeah
used to have a alix/wrap at the base of each wifi tower, running squid in transparent mode for WISP clients.
err s/wifi/wireless/
90 gallon saltwater
and, wrong window. RandalSchwartz: heh
saltwater NFS mounted, was it? jpalmer: lol yup. growing corals in my fbsd machine :P RandalSchwartz: how well does NFS work with fbsd?
could I do that mounting here? jpalmer: ehh, I'm not an NFS fan. so, I literally have never used it for anything else. RandalSchwartz: ok jpalmer: you should be able to.. as long as they are on the same VLAN. export shares on the buildworld machine, and mount them on the installworl machines RandalSchwartz: interesting jpalmer: or if you have some of the fuse modules, you might be able to do it with sshfs or something instead.
does rsync honor the extended attributes, like the immutable flag and such?
I'd make sure it does, before trying to use rsync to clone /usr/obj and /usr/src RandalSchwartz: rebooting first machine onto new kernel jpalmer: up_the_irons: do you happen to track or poll the vps's, to see which versions of fbsd people are running? up_the_irons: jpalmer: not really jpalmer: ahh, was just idly curious of how quickly people move their vps's to the latest versions after release.
no worries up_the_irons: jpalmer: i get a small flood of requests for iso changes after a release, but nothing spectacular. i think most people just stay with whatever they got in the beginning jpalmer: gotcha. I was wondering if you building world on a machine.. and then making it available to clients for installworld.. if that'd be a sinificant "benefit" to a large part of your client base. but, doesn't sound like that'd be the case RandalSchwartz: o/~ it's install world, after all... it's install world...after all..." up_the_irons: jpalmer: that might be a cool add-on, yeah jpalmer: up_the_irons: if a lot of people would make use of it, sure. but just based on what you described.. Not sure i'd be much of a valueadd to the majority
ie, prolly more effort and headache than it's worth. up_the_irons: perhaps jpalmer: I think a centralized, and well tuned database server.. or backup server would probably be more useful to people. jdoe: jpalmer: dunno if it's really worth it. jpalmer: which? or all of the above? jdoe: jpalmer: anyone who's still using buildworld instead of the binary updates is likely doing it because they have updates.
er
customizations jpalmer: jdoe: that's a good point. very true. RandalSchwartz: or - we've been burned by binary updates (merge3 instead of mergemaster!) and we now want a clean upgrade
the last time I tried a binary update, merge3 wanted me to manually merge the CVS tags in 150 files
No. f'ing. way.
so I'll never do a binary update again
with mergemaster, I can ignore the cvs tags, and it's about 5 minutes per machine
that was also the consensus of the fbsd mailing list jpalmer: I stopped using fbsd a few months after I found freebsd-update. but the few times I *did* use it, it seemed to work pretty well up_the_irons: i am surprised at the prevalence of fbsd people compiling their own stuff. don't get me wrong, i will compile my own stuff when no binary package exists, but otherwise, on ubuntu, it's all about apt-get
aptitude safe-upgrade, is just too easy jdoe: RandalSchwartz: ymmv. I've never had a problem, other than how ungodly slow it can be to apply 100000 patches. jpalmer: up_the_irons: when I was primarily in fbsd world, I actually enjoyed buildworld/installworld. there is just something.. soothing about watching the screen scroll by compiling fresh clean sources.
But, when I started this job and decided to learn RH by immersion learning.. I gotta say, I'm starting to like it pretty well. yum update is pretty damned nifty. up_the_irons: LOL
jpalmer: i shun RH b/c of how many times i got hacked in the past. after i started using debian / ubuntu, never got hacked again. but i hear what you're saying. when i've had to use RH or Cent, yum is pretty convenient jdoe: jpalmer: I only have two complaints about rh/centos ... one is that it has too many shoot-yourself-in-the-foot commands like sys-unconfig... the other is that major version upgrades aren't supported at all. jpalmer: jdoe: major version upgrades aren't *officially* supported by most vendors, including freebsd :)
doesn't mean they aren't possible jdoe: they're not? FreeBSD gave explicit instructions on how to upgrade 8 to 9. And 7 to 8. And 6 to 7 :P
(and deb/ubuntu support it as well) jpalmer: instructions, yes. but ask them, it's not officially supported jdoe: centos, 4 to 5 was possible (with some hackery), 5 to 6 is also... possible, but wildly unpleasant. jpalmer: I tried fbsd 4 to 5, on multiple occasions. never had one go successfully.
I think it was 4 to 5, back when they changed udev to devfs, or whatever the big thing was. jdoe: worked for me, but I had some packages I had to clean by hand.
re: freebsd, dunno what constitutes "official support", but it's mentioned in the handbook.
also, my centos systems are relatively bare (java shop)
so less to go wrong. jpalmer: *nod* I tend to keep all of my machines specific to one task/purpose these days. -: RandalSchwartz types csup on 4 machines RandalSchwartz: I think this time, I'll just do what I know works. :)
next time, I'll try the rsync jpalmer: a lot easier to keep them updated, or do big updates, if you only have to worry about 1 service or application. RandalSchwartz: make buildworld buildkernel on three of four machines now
earning my disk I/O's tonight. :) up_the_irons: love full version upgrades in deb / ubuntu :) ***: Hien has quit IRC (Read error: Connection reset by peer)
Hien has joined #arpnetworks
HighJinx has quit IRC (Quit: Computer has gone to sleep.) jdoe: up_the_irons: we're apparently switching at work, nobody knew until now that major versions are a mess in centos. -: jdoe wonders how pleasant that will be.
jdoe also can't wait for debian-openssl 2.0 up_the_irons: jdoe: hah
jdoe: i didn't know either, until u guys mentioned it jdoe: yeah. I figured the 4->5 path was typical (ie "awkward, but you can do it")
... but apparently not. up_the_irons: hah jdoe: and we're not big/rich enough to be able to reinstall/replace trivially :P up_the_irons: i hear ya :) jpalmer: jdoe: maybe it's time to go with puppet and theforeman :P
reprovision a machine, and have it rebuild the appstack in automated fashion :P
and now that they've integrated libvirt, and made it painlessly provision KVM instances.. I'm all about it! up_the_irons: they integrated libvirt? nice ***: HighJinx has joined #arpnetworks jpalmer: yeah. not released yet, but you can check it out via git.
actuall, as much as you use KVM, and as often as you provision hosts.. you might want to look into it. with theforeman, you may be able to have your billing system kick off a VM provisioning profile.
for new signups, that is. ***: HighJinx_ has joined #arpnetworks
HighJinx has quit IRC (Remote host closed the connection)
HighJinx_ is now known as HighJinx
HighJinx has quit IRC (Ping timeout: 252 seconds) kraigu: jdoe: I had a FreeBSD box I took from 5 to 6 to 7 (make buildworld style), no worries
another one went from 6 to 7 to 8 (and 8.1 to 8.3 last night) -: kraigu chimes in very late ***: HighJinx has joined #arpnetworks jdoe: nobody asked you >:|
haha kraigu: nobody said not to speak either =]
and I just found out a 4.6 box I set up in 2002 is... still running, although it's at least been upgraded to 4.11 in the interim
simultaneously pleased with my purchase and setup, and despairing of the guy who took it over after I left in 2003 :( jpalmer: kraigu: I have half a dozen 2.2.6 machines still in production kraigu: I don't know if that's something to be proud of :(
(do you mean freebsd 2.2 or linux kernel 2.2?) jpalmer: freebsd 2.2.6
and why wouldn't I be proud of that? that's a long damn time in production. and it just keeps running.
rock solid. kraigu: is it internet-facing? jpalmer: do you think I'm a moron? lol kraigu: mine is :[22:06] <jpalmer> these are point of sale machines. not internet connected.
well, the one that was mine 9 years ago, but has since passed into other hands
runs web+mailservices, completely open.. sigh jpalmer: the only time these machines get upgraded, is when the physical hardware dies. kraigu: I just can't believe the hardware's still fine too
it wasn't anything special, just off-the-shelf PC hardware. fuji 10GB hard disk, IIRC. jpalmer: yeah, it's amazin how well-built older hardware is. kraigu: well, the fuji 6.4s at the time were complete crap
but we put a BIG DISK in it =] jpalmer: exceptions prove the rules ;) kraigu: something like that
anyway, point was binary or source upgrades of freebsd, major release or no, is certainly possible
I have a hate-on for freebsd-upgrade after last night though, mergemaster took a step backwards!
HEY GUYS LOOKS LIKE THE CVS ID CHANGED ON THIS FILE, CARE TO CHECK OVER MERGING IT?! -: kraigu throws up jdoe: kraigu: we just noticed today our prod nis (lol) servers have uptimes of ~1300 days. kraigu: jdoe: nis, enough said =] jdoe: yeah. We wanted to deco them, but now we feel bad about it.
and there's been vague talk about trying to hot-splice a ups onto it while the health and safety guy is out of earshot... kraigu: kill your babies man
before they eat you
learn from cronus! :D jdoe: not mine, alas... my best is ~600 :( kraigu: tiny babby!
well, if they decide to do the UPS thing, you need a video of it, the internets demand it jdoe: haha.
if they decide to do the UPS thing, I'll be the guy with the wooden broomstick prying one of them off the cage :P kraigu: haha
and the video! ***: amdprophet has quit IRC (Quit: Leaving...)
amdprophet has joined #arpnetworks