***: mike-burns has quit IRC (Quit: WeeChat 0.3.7)
chmod has quit IRC (Ping timeout: 246 seconds)
chmod has joined #arpnetworks
chmod has quit IRC (Ping timeout: 246 seconds)
chmod has joined #arpnetworks
Webhostbudd has quit IRC (Quit: Leaving)
LT has joined #arpnetworks
Ehtyar has joined #arpnetworks
Mexicainvexed has quit IRC (Remote host closed the connection)
dzup has joined #arpnetworks CaZe: qbit: I refined my alignment technique again. :D qbit: haha
nice CaZe: Basically, you don't have to tap the burr at all.
Starting with all the bolts loose, if you rotate the axle, it shouldn't bob, because the outer burr is free to slide around to accomodate the inner burr.
Now if you tighten one bolt all the way, it will probably start bobbing.
Rotate the axle and stop when it's reached its highest point.
Now go back to the bolt you just tightened, and loosen it all the way, while simultaneously pressing down on the top of the axle.
The axle should drop down as you loosen the bolt. qbit: makes sense CaZe: Now tighten the bolt again.
The axle won't move back up as you tighten the bolt.
Now see if it still bobs.
If it does, repeat.
If it doesn't, then move on to the next bolt and perform the procedure. qbit: cool CaZe: With practice, you should be able to tighten-loosen-tighten each bolt once. qbit: i will give it a shot sometime today CaZe: I think it works better if you take out the springs.
I put the small bolts back in though.
Without the springs. qbit: to help with keeping the outerburr in place? CaZe: No, tighening the three long bolts will keep it in place.
It's to keep the little clear acrylic cylinder in place while you're aligning. qbit: i just mean when it's mostly in pieces
right CaZe: The first time I did the above procedure, after I finished, I saw that the cylinder had moved so far to the side that there was a gap at the bean loading opening. qbit: whoa
heh CaZe: I keep the little bolts loose, of course.
I should make a video of this and send it to Doug.
I wonder if anyone else has developed this same technique by now, though. qbit: i am a fan of over information
:D
gonna do it? CaZe: Make a video?
Maybe, I'd have to clean up.
I think I'd actually need a cameraman though.
THe shot would have to zoom in on the top of the axle to see the bob. qbit: kew CaZe: Maybe I'll just describe it to Doug, and he can make a video of it with Barb. qbit: she is pretty pro at camerain ***: mike-burns has joined #arpnetworks
ChanServ sets mode: +o mike-burns up_the_irons: brycec: can you rsync with amazon S3 RSS? last i used S3, it was a pain to use with filesystems, especially large files. i always had reliability issues (backups stalling in the middle and nothing telling me why). If S3 RRS works for you, I would say keep using it and there is no need to buy or use any local rsync-based backup storage. It's just that many people have requested local backup
storage, so i'm trying to find the best way to provide it
CaZe: free dns services drop your IP for inactivity?
mercutio: if you only had 2 disks, why would you got with raidz or mirror? (i'm new to this, I've worked recently with ZFS mirrors, but not raidz yet)
dang, lots of scrollback, i'll read the rest later... CaZe: up_the_irons: The ones I've tried do. dyndnds, afraid.
*dyndns up_the_irons: what about HE? CaZe: HE? RandalSchwartz: hurricane electric CaZe: Do they provide domains or just DNS hosting? RandalSchwartz: I think just secondary
for their ipv6 project
I have them for stonehenge.com secondary
HE provides connectivity
but they are also getting people aware of ipv6
weird. ipv6.he.net is blocked by opendns for having viruses. :) ***: heavysixer has joined #arpnetworks
ChanServ sets mode: +o heavysixer
chmod has quit IRC (Ping timeout: 246 seconds) brycec: RandalSchwartz: is tunnelbroker.coM? RandalSchwartz: yeah that brycec: CaZe: HE supports dynamic IP's on domains they're hosting, dns.he.net. But you do have to bring your own domain for that to work.
However, HE's free DNS is definitely my favourite. (And contrary to popular belief - there's nothing special about it, no strings attached, it's not ipv6 only, yadda yadda yadda)
up_the_irons: Quite right, no rsync. But rsync isn't really the best for doing backups. Simplest perhaps, but it leaves thigns to be desired compared to more "big boy" backup solutions that provide manifests, incremental/staged (sure, there's rsnapshot), and that sort of thing.
up_the_irons: FWIW I use duplicity+s3 and it works fantastically. Everything is stored encrypted so I'm not worrying about some rogue Amazon employee stealing my data through backups. It's got a slew of features to handle incremental, full-if-older-than, cleanup, verification, and yes manifests. And it just throws the data at S3 and it sticks :p
I won't say S3 is perfect - if you've ever tried to read back data you just pushed, you understand the frustration. But like you said, up_the_irons, it's working for me and that's what I'll probably continue to use. up_the_irons: brycec: that sounds pretty cool brycec: But as a customer, I thank you for keeping your eyes out for customer demand and working to provide us with that. Hopefully you'll get some sales bullet-points from this.
(It's easy! it's simple! It's fast!) up_the_irons: hahaha
yeah RandalSchwartz: did you look at tarsnap as well as duplicity ***: gcw|mbpro has joined #arpnetworks brycec: up_the_irons: You'd raidz two disks to get capacity. It still does its parity thing, you can adjust whether some data gets duplicated more than others and all the other ZFS-isms, but you are still at risk of losing *some* data if a drive fails. (but hey, zfs will tell you what files you've lost ;)) RandalSchwartz: raidz2 4 disks would be better than raidz + mirror though brycec: RandalSchwartz: I have... And I want to like and use it, but I can't justify their pricing in comparison. RandalSchwartz: what pricing? I thought it was pretty much S3 passed through
with compression and de-dup done client-side brycec: RandalSchwartz: S3 b/w is free to store data, S3 storage is $.09/GB/mo for RRS (and I think $.12/GB/mo for standard). Tarsnap is $.30/GB/mo, plus b/w
Oh and duplicity does compression and some dedupe too (not positive to what degree off the top of my head) RandalSchwartz: the original "2 data centers lost" version of S3?
or the cheap version
that might be part of the difference brycec: RandalSchwartz: cheap version is RRS - reduced redundancy RandalSchwartz: yeah - so you're getting less service brycec: I'm paying for a lesser committment, yes RandalSchwartz: thus I'd expect a pricing difference.
ok - just working this out brycec: RandalSchwartz: but for the "standard" S3, Amazon's price is $.125/GB/mo
and tarsnap is $.30/GB/mo
I don't begrudge him his overhead or anything. Like I said, it seems like a great product, but it's not what I use because I don't want to spend that much. RandalSchwartz: .. http://www.daemonology.net/blog/2008-12-14-how-tarsnap-uses-aws.html brycec: (I know THAT he uses S3, yes) RandalSchwartz: actually - that's probably a recent price decrease, and he hasn't adjusted to follow suit
he says $0.15/GB there
plus bandwidth brycec: http://aws.amazon.com/s3/#pricing
aaand I'm out ***: chmod has joined #arpnetworks RandalSchwartz: ... http://www.zdnet.com/blog/diy-it/amazon-reduces-s3-prices-because-0-11-is-too-much-to-charge/400
yeah, reduced ***: cullum has quit IRC (Read error: Connection reset by peer)
cullum has joined #arpnetworks toddf: if you want an archival / backup product that doesn't charge for bandwidth, see cyphertite.com ;-) (disclaimer: I'm only a happy customer, not even getting perks for referrals.. yet..) RandalSchwartz: ahh - they use a datacenter, and not S3
wondering how they got to $0.10/GB/mo
and somehow, they are magically on CST even though the rest of the middle of the country is on CDT. :)
<<== timezone nerd qbit: so for userscripts
do i have to do something to enable them?
:userscripts list it
but it doesn't seem to be working
i do have script whitelisting
but "enable-scripts" is set to true
lol - wrong channel
^^^ luakit for context ***: LT has quit IRC (Quit: Leaving) jdoe: RandalSchwartz: I'm confused how the guy in that article got to 0.11, unless he's using reduced redundancy... which seems odd for backups. ***: chmod2 has joined #arpnetworks
chmod has quit IRC (Ping timeout: 246 seconds) brycec: lol qbit fail qbit: :D brycec: Also cyphertite is pretty nifty. I've played with it in the past, no problems with it... but it wasn't my style, a bit too complicated.
jdoe: I wouldn't say there's anything wrong with using RRS for backups. It's still highly reliable, just not as highly available. Amazon's still got two copies of the data. qbit: ^ that's the prob i am having lately
having multiple hosts is kidna a bitch brycec: Not to mention the chances of you losing your primary data the same time as Amazon losing your backup are pretty damn small. qbit: especially if you don't want host a to be able to see host b's backups brycec: qbit: heh I banged my head against the AWS/IAM wall for a few weeks trying to get that to work. For now, I just hope my clients don't look at the script and see my access keys. Which... I'm pretty safe on, but still worry about. Lefty: rot13 them qbit: that's lame :P
haha brycec: qbit: I think you mean ynzr :P
I really should make rot13.com my homepage qbit: yeah you should brycec: Side-note: If this counts for anything, duplicity is what Ubuntu uses for their in-built backup solution. (Well, deja-dup which uses duplicity) jdoe: brycec: 99.99% isn't highly reliable when it comes to backups. andol: availability or durability? brycec: reliability = durability, availability = availability jdoe: andol: RRS = 99.99% durable = 1/10000 loss in a year.
(on average)
RRS is fine if you want a ghetto CDN or something, probably not ideal for backups. andol: jdoe: Aware of the implications, was just wondering what the 99.99% was refering to, without seeing that from a quick glance as at the backlogg. jdoe: IMNSHOIANABBQ brycec: ah ***: chmod2 has quit IRC (Ping timeout: 246 seconds)
chmod has joined #arpnetworks
chmod2 has joined #arpnetworks
HighJinx has quit IRC (Quit: Computer has gone to sleep.)
chmod has quit IRC (Ping timeout: 246 seconds)
Webhostbudd has joined #arpnetworks
chmod2 has quit IRC (Ping timeout: 246 seconds)
HighJinx has joined #arpnetworks
chmod has joined #arpnetworks
cmeiklejohn has joined #arpnetworks qbit: CaZe: i forgot to ask -did you use washers in your new method? ***: Ehtyar has quit IRC (Quit: IRC is just multiplayer notepad)
gcw|mbpro has quit IRC (Remote host closed the connection) mercutio: up_the_irons: oh i meant go for an extra disk, but if that isn't an option then mirroring probably isn't an option either CaZe: qbit: Nah. qbit: kew CaZe: I just kind of held the nut over on one side. toddf: up_the_irons: late to the party but .. I note you inquired about rsync backup service. If you wanted us to all play, even if we don't have the budget to pay, perhaps offer limited accounts with limited disk space, and charge for extra space. I know I could probably backup important config files and the like in < 128mb or even < 1gb and you might even get some people who decide to pay once they reach the level of data backup that requires them to .. just mercutio: 128mb disk space
heh ***: cmeiklejohn has quit IRC (Quit: ["Textual IRC Client: www.textualapp.com"]) toddf: 0$ cd /
0$ sudo tar cf - etc | gzip -9 | wc -c
1487236
if I want just the config preserved, I can do it trivially. data of course is going to take more than 1gb so I will decide if I want to pay for data or just the configs .. though one could easily store that small archive anywhere, not just a free rsync account mercutio: hmm
# tar cf - etc | gzip -9 | wc -c
1615487
# tar cf - etc | xz -9c | wc -c
1005208
you could just email something that size
and fwiw xz -9c is really slow toddf: and takes a bunch of mem too
xz -7c is about the max I've found useful, unless you want to archive stuff in the bg on an archive server that has nothing better to do but recompress at higher compression ratios mercutio: well for write once read many times it may be helpful
gah this linux is pissing me off
on my desktop ***: Ehtyar has joined #arpnetworks mercutio: it seems like gnome-terminal crashed, like surely that's something that should never crash
either that or tar crashed in the midle of outputting a filename
oh sweet! resizing fixed it
i bet /etc is bigger on linux
# tar -cf - etc | gzip -9c | wc -c
2862681
twice your size :) toddf: thats linux's problem not mine *grin* mercutio: it's a desktop too so may have more things in /etc
oh what du says X11 is 65mb big
oh there's a core dump in /etc/X11
now it's normal size
# tar -cf - etc | gzip -9c | wc -c
1588074
that's actually smaller than my openbsd vps
and xz was way faster
i don't think xz is necessarily too slow for real use
it seems 1/4 of the speed of gzip toddf: xz -9 can be. mercutio: yeh with bigger working set maybe
i'll try -6
-6 is the same toddf: lets say you have streaming backups to a backup system using xz -9 will be the bottleneck. using xz -1 and then later when no backups are running maybe consider using xz -9 but *shrug* mercutio: err -6 is slightly faster on arp the same on my desktop toddf: I've specifically chosen gzip as a 'not as compact but not holding up the backups' algorithm before mercutio: but if you're going over the internet to an adsl modem xz -9c should go faster than adsl speeds shouldn't it? toddf: for less data xfer and/or less churn, xz -7 is usually where I go
it is implementation specific, lots of factors, including cpu speed, and adsl plan mercutio: oh actually i'm doing 500k/sec
heh
-3c with lzma is faster than modem speeds while being better compression than gzip
this is only on single core it seems too toddf: "implementation specific" indeed ;-) mercutio: well adsl is generally limited to 24 megabit maximum
so if you get 3megabytes/sec it should hold true all the time
but i don't really hear of people getting anym ore than 2.5 megabytes/sec
with uploading on the other hand on adsl -9 would make sense :)
people have been compalining on an online forum recently how hard it is upload data overseas
one guy was considering mailing a usb stick
(he wanted to upload 250gb) toddf: adsl needs to die. its not keeping up. unfortuantely for some it is the only option. thankfully I'm in Oklahoma City where COX has overbuilt their fiber and I got a decent speed inet connection. mercutio: i have adsl at home
it's the upload speed that's annoying
i kind of wish there was an adsl 3 or something
that always included annex-m as an option
and did better dynamic bit loading
cos adsl is going to be around another 10 years probably ***: gcw|mbpro has joined #arpnetworks mercutio: toddf: xz -1c is faster than gzip -9 on both vps, and desktop
and gives higher compression ratio
so maybe xz -1c is useful for fast compression now over gzip toddf: you cannot make a general rule
you can simply build a speed chart of various algorithms vs typical compression ratios
and decide where the sweet spot is for your given application
also consider xz takes more mem than gzip by quite a bit
if you have many xz running in parallel or a low mem system that may also factor into the equation mercutio: ahh yeh cache misses etc toddf: once when I was nieve I set 'xz -9' and ran into swap on a system that was doing multiple backups in parallel
not pretty mercutio: running into swap is never pretty :) jlgaddis: $ du -hs vps-backups/
860M vps-backups/
that's all of my databases (mysqldump'd) and /var/www tar'd up (not compressed). i'm surprised it's that small. up_the_irons: mercutio: going for an extra disk isn't an option cuz the blade server only has 2 disks. i'm fine with mirroring at this point. toddf: iscsi! heh up_the_irons: toddf: yeah, thanks for the tip mercutio: oh right, so you'd need to migrate to a diff server if you wanted to create more space up_the_irons: FreeBSD question -- Do anyone of you guys have SSH set up as chroot on FreeBSD?
CaZe: it has been on my mind to offer forward DNS services ( CaZe │ Those free dns services are a little annoying...)
CaZe: just other things keep taking priority.. the way i set up my reverse dns manager, with the powerdns models i wrote, it shouldn't be *too* hard to implement
Webhostbudd: not sure about freebsd, but toddf and mercutio finally got virtio disk / NIC working on OpenBSD (on the new server) mercutio: up_the_irons: oh i was finding an issue with network latency when doing heavy disk i/o btw
sometime after you said to benchmark more :)
it's quite reproducable up_the_irons: CaZe: qbit : what is it that you guys keep talking about, with the bols, nuts, alignment, etc... ?? :) mercutio: basically sh -c 'sync && tar vxpzf /root/base52.tgz && sync' while leaving a ping running to it mjp: up_the_irons: nope.. shell users get put in a jail or i use vsftpd+ssl if they need ftp access. qbit: up_the_irons: coffee grinder: http://www.orphanespresso.com/OE-PHAROS-Hand-Coffee-Grinder_ep_636-1.html up_the_irons: brycec: about raidz on two disks -- losing data, even if only some, isn't an option; just don't want that headache mercutio: i have no idea if it's the disk or network driver causing it though, but vmstat shows hardly any interrupts when there's lag up_the_irons: i just want to pop in a new disk and rebuild mercutio: up_the_irons: you can't raidz with two disks, i thought you may be looking at expand later when you said starting with 2 disks
but i see expand means san or hardware migration now :) up_the_irons: mjp: roger
qbit: roger
qbit: you're really into tuning that thing... :) qbit: CaZe is mroe into it than I :D - he has theories even! up_the_irons: mercutio: yeah i would just hardware migrate it. pop the two disks into a pretty much identical server, but with more disk slots
qbit: LOL brycec: up_the_irons: Yeah I know it's unwanted, particularly in this case. I was just answering the question of why anyone would bother raidz two disks. CaZe: earthquake up_the_irons: toddf: are you using iscsi for production stuff? I looked into it a few weeks ago, but didn't come away with anything great. granted, i only looked at Linux solutions. What do the OpenBSD solutions look like?
brycec: ah ok CaZe: In Boston! mercutio: gah, seagate 3 tb disk connecting at sata 1 speed up_the_irons: first world problem mercutio: yeah up_the_irons: ;) mercutio: i don't know if it's something worse
err a symptom of something worse -: milki is perfectly happy with dma33 >.> mercutio: otherwise it doesn't relaly matter cos it's just for archiving stuff
i'm a bit weary cos it's a forward replacement for rma
oh, it's showing up as 2tb only too
Sector size (logical/physical): 512 bytes / 4096 bytes
I/O size (minimum/optimal): 4096 bytes / 4096 bytes
up_the_irons: that's that 4k sector thing i was mentioning to you before up_the_irons: mercutio: i c mercutio: oh, you have to use gpt!
i didn't realise standard partition tables didn't work with 3tb
i thought it was just a bios issue
i was using gpt previously
(and zfs)
actually it may have had no parititon table for all i know up_the_irons: this is specific to Linux, but I don't see why it wouldn't work for FreeBSD: http://www.techrepublic.com/blog/opensource/chroot-users-with-openssh-an-easier-way-to-confine-users-to-their-home-directories/229
any FreeBSD users care to weigh in on that? mercutio: i think openssh has some built in stuff
i think some of thechrooting stuff was used for cvs ages ago up_the_irons: mercutio: yep, > 2TB is one of the advantages of GPT mercutio: this hard-disk is way slower than ssd :)
but at least it's not really noisy when just doing an rsync milki: *bsd systems have jails which probably work better than chroot mercutio: mhttp://olivier.sessink.nl/jailkit/howtos_cvs_only.html
something like that may help you
you also should disable forwarding X11, forwarding ports etc
http://www.thegeekstuff.com/2012/03/chroot-sftp-setup/
stuff like that may help too
but i imagine people want to run rsync
and tar and stuff
steadfast give people shells on their backup server fwiw
i think it was linux with remote storage mnathani: up_the_irons: are you considering block size options (5gb, 10gb) etc for the local rsync backup or pay per usage pricing model? up_the_irons: mnathani: i was thinking 20 cents per GB, minimum 5GB ($1)
milki: if i'm just giving shell accounts for rsync, et. al. backups, i am under the misimpression that jails are too much? i would think chroot would be enough...
mercutio: i always disable all the forwarding stuff milki: ya, probably too much -.-
though for project classes, we did do a debian vserver per student -.- up_the_irons: i'm now a *paying* user of PagerDuty
milki: nice
i remember getting a Solaris shell account when I was at UCLA
or maybe it was still SunOS back then, don't remember... milki: i dont think we have solaris machines anymore
probably when the licensing terms changed... up_the_irons: grad students got Solaris *desktops*
i thought that was particularly pimp milki: lol
i remember labs full of solaris desktops and dumb terminals -: milki shudders milki: hated those dumb terminals up_the_irons: someone tell me there is an IPv6 capable monitoring service (like Pingdom)
i will sign up right now...
milki: hah milki: default de was cde >.< up_the_irons: yup :) mercutio: maybe you should start one up_the_irons :) up_the_irons: I will if heavysixer wants to help me -: up_the_irons looks at heavysixer mnathani: PagerDuty's infrastructure is fully replicated in two different Amazon AWS data centers << nice up_the_irons: sweet
channel poll -- useful to anyone? http://stats.pingdom.com/31n57wtu0gs4
(i only have the 5 server plan, with 4 hosts configured)
someone had asked for a public status page, so...
w00t!! http://blog.uptimerobot.com/new-features-bulk-actions-ipv6-support/
ipv6! -: up_the_irons signs up jbergstroem: up_the_irons: can't find a pricing page. you? up_the_irons: jbergstroem: it's free actually jbergstroem: ah, so 50 is the limit. period. up_the_irons: jbergstroem: until Feb 2013 they say ***: gcw|mbpro has quit IRC (Ping timeout: 260 seconds) jbergstroem: ok ***: Mexicainvexed has joined #arpnetworks
dzup has quit IRC (Read error: Connection reset by peer)
gcw|mbpro has joined #arpnetworks
dzup has joined #arpnetworks
Mexicainvexed has quit IRC (Read error: Connection reset by peer)
jlgaddis has quit IRC (Ping timeout: 240 seconds)
gcw|mbpro has quit IRC (Remote host closed the connection)
Mexicainvexed has joined #arpnetworks
gcw|mbpro has joined #arpnetworks
dzup has quit IRC (Read error: Connection reset by peer) jbergstroem: up_the_irons: thansk for hte pingdom link. good to have for comparision
-i ***: jpalmer has joined #arpnetworks
jpalmer has quit IRC (Client Quit)
ryk has quit IRC (Ping timeout: 248 seconds)
ryk has joined #arpnetworks
ryk has quit IRC (Changing host)
ryk has joined #arpnetworks
staticsafe has quit IRC (*.net *.split)
er|c has quit IRC (*.net *.split)
staticsafe has joined #arpnetworks
er|c has joined #arpnetworks up_the_irons: jbergstroem: np
jbergstroem: do u mean the uptimerobot link? phlux: interesting
does that pingdom stuff work with multiple servers? up_the_irons: phlux: what do u mean? ***: er|c has quit IRC (Ping timeout: 276 seconds) up_the_irons: jbergstroem: or do you mean my public status page on pingdom? phlux: would i need separate accounts to check the uptime of all of my vpses? up_the_irons: phlux: if you sign up for pingdom?
phlux: i'm on the 5 hosts plan, so i can check up to 5 hosts phlux: ah ok ***: jpalmer has joined #arpnetworks jpalmer: up_the_irons: ping ***: gcw|mbpro has quit IRC (Remote host closed the connection)
HighJinx has quit IRC (Quit: Computer has gone to sleep.)
er|c has joined #arpnetworks jbergstroem: up_the_irons: the public status page up_the_irons: jbergstroem: np, it's not really official... i mean there are only 4 hosts
jbergstroem: but it will do the job for those 4 hosts :)
FreeBSD peeps, i need help --
mfsbsd# dmesg|grep igb
igb0: <Intel(R) PRO/1000 Network Connection version - 2.2.5> port 0xe020-0xe03f mem 0xf7980000-0xf79fffff,0xf7a04000-0xf7a07fff irq 16 at device 0.0 on pci1
igb0: Using MSIX interrupts with 9 vectors
...
How do I find out the model of that NIC?
like, the exact model... jpalmer: pciconf -v IIRC up_the_irons: and, is the 2.2.5 driver the official Intel one? (cuz I don't see it on their site, although I see a 2.2.3 one) ***: gcw|mbpro has joined #arpnetworks jpalmer: pciconf -v will give you a vendor and hardware number. then you can go to pcidatabase.com or equiv and get exact model number up_the_irons: ah cool jpalmer: let me know if that works. been emersed in CentOS for going on 3 years now. my BSD-fu is rusty mercutio: it's probably an i350 i'm guessing
so apparently the net in general has issues tonight
common thread seems to be above.net having issues up_the_irons: jpalmer: "-v" didn't work, but pciconf by itself gave some info. looked up stuff, said "SuperMicro". it is indeed a supermicro server, but could not find NIC. Found one "close". In the end, I just did a MB lookup and found the NIC (82580DB) mercutio: up_the_irons: probably cos it's integrated up_the_irons: mercutio: yup jpalmer: up_the_irons: maybe it was pciconf -lv?
(try it and lemme know) up_the_irons: if I ssh into this new FreeBSD box (bare metal, not vps), and cat a large text file, it freezes. I had this issue on kvr27 too (on Linux), before I updated the igb driver. The FreeBSD box also uses its own igb driver. mercutio: i still don't understand how bgp is meant to work when providers accept traffic and don't forward it up_the_irons: jpalmer: w00t, nice "-lv" does it :) mercutio: and it seems to usually happen to "in the middle" providers jpalmer: ahh, thats good to know. I was like.. I KNOW it's pciconf! up_the_irons: mercutio: yeah, host reachability is an issue
mercutio: a bad peer will make your life hard mercutio: 4. xe-2-3-0.cr1.lax112.us.above.net 0.3% 327 1.4 4.6 1.1 62.6 8.2
5. xe-4-1-0.cr1.iah1.us.above.net 46.3% 327 35.6 38.1 32.8 83.9 8.9
that's from arp
but i tried from our network as well, after having total outage to a site shifting to different outbound
and it still routed through above.net but worked
just higher latency
it seems mzima have a connection to above.net
but i don;'t know how many above.net hosts are effected
therre really should be health advertised through
ok finding random route from bgp table shows it's more than one above.net destination ***: HighJinx has joined #arpnetworks mnathani: Anyone else experiencing UDP Packetloss resulting from DNS Queries originating at an ARP VPS? mercutio: not that i've noticed
maybe try iperf at 2 megabit?
iperf 2 megabit udp shows no packet loss for me mnathani: mercutio: I would appreciate if you could try the following DNS query: dig @ns1.timewarner.net cnn.com mercutio: it's going slow
whatever it is
# dig @ns1.timewarner.net cnn.com
; <<>> DiG 9.4.2-P2 <<>> @ns1.timewarner.net cnn.com
; (1 server found)
;; global options: printcmd
;; connection timed out; no servers could be reached
;; Query time: 29 msec
;; SERVER: 204.74.108.238#53(204.74.108.238)
from nz it was instant mnathani: The same query works just fine from Washington as well as London UK mercutio: but strangely so
is ns1.timewarner.net anycast? mnathani: not sure mercutio: oh it is anycast
it also doesn't respond to pings
oh look
od a mtr
12. ONE-SOURCE.edge1.SanJose3.Level3.net 85.7% 8 9.4 9.4 9.4 9.4 0.0
that'll be the issue
umm
http://internethealthreport.com/Main.aspx?Destination=Level3
look at that
it seems level 3 have had issues recently
anyway, they should have more than one nameserver on different networks even if using anycast
and they don't
so it's timewarner's fault
as an interim measure you can use a secondary dns server on a different network
although that doens't mean the site will work mnathani: mercutio: Thanks for looking into this mercutio: <root@arp:~>
zsh 3016 # time curl http://www.cnn.com/ > /dev/null
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
100 127k 0 127k 0 0 1628k 0 --:--:-- --:--:-- --:--:-- 1750k
but it is working in this instance mnathani: I was testing a set of web tools for working with DNS mercutio: ok
anycast can be slightly confusing toddf: up_the_irons: wrt iscsi, I use netbsd-iscsi-target for stuff in my home office, I wouldn't recommend it for anything but iscsi initiators that can handle reconnecting upon occasion, you have to restart it to reconfig anything.. wrt a client who paid way to much, we have an equalogics array, talk about scalable in a big way, they can cluster the storage and even migrate 15mb chunks of disk into fast disks on different (independent) disk shelves and have up_the_irons: man, how do I load a driver in FreeBSD that's already included in the kernel?
mfsbsd# kldstat
Id Refs Address Size Name mercutio: up_the_irons: kldload? up_the_irons: 1 8 0xffffffff80100000 e56d68 kernel
2 1 0xffffffff8227d000 11b28 ahci.ko
3 1 0xffffffff82412000 4d6e tmpfs.ko
mfsbsd# kldload ./if_igb.ko mercutio: oh up_the_irons: kldload: can't load ./if_igb.ko: File exists
mfsbsd# mercutio: you mean you want to unload it and then reload it
oh
you don't compile it in up_the_irons: mfsbsd is a read-only ISO
so i can't change it... mercutio: oh up_the_irons: man this project just keeps getting more fucked... mercutio: easy solution
stick a pro1000 card in temp -: up_the_irons listens mercutio: igb can cause issues for some people
pro1000 is em driver and used to cause issues for some people
but it's been around a long time and pretty stable now up_the_irons: that's not real easy when the server isn't next to me ;) mercutio: oh
true that
uhh up_the_irons: any kernel boot param I can use to prevent igb from loading? mercutio: and it's not easy to mount a different iso? up_the_irons: i *can* mount a different ISO, yeah... mercutio: maybe try 9.1-rc2? up_the_irons: maybe i need to go back to the LiveCD mercutio: it'll be stable soon anyway up_the_irons: yeah..
i'll come back to this later -: up_the_irons wanders off mercutio: heh
aren't os installs fun
mnathani: that timwarner dns server is working again on arp btw ***: Webhostbudd has quit IRC (Read error: Connection reset by peer)
Webhostbudd has joined #arpnetworks toddf: up_the_irons: chroot with sshd is how I've setup chroot envs before on OpenBSD. works great! I've even done chroot for sftp for uploading webstuffz...
since chroot is a unix technique I can' see why it would not work on all flavors of unix mnathani: mercutio: Thanks
It is working fine now: http://dns.winvive.com/dns.php?Domain=cnn.com&QueryType=A mercutio: you may want to use a secondary server to test doing the queries from too staticsafe: anyone else notice some weirdness on the internet (connection issues, dns failures) lately? mnathani: The link above tries to find each authoritative server and tests each one mercutio: static: yes :)
not a lot
but there's the issue mnathani raised with dns server not responding, and there was an issue with above.net having 50% packet loss staticsafe: ouch mercutio: well actually it was 100% packet loss when i first noticed it
the above net issue seemed to effect traffic from los angeles to dallas at least
and mnathani's issue with dns was with someone hosting multiple dns servers in one location, rather than spreading them out
well they were spreading htem out by using anycast, but each name serrver mapped to the same-not-working-location staticsafe: hmm yea mercutio: which i'm sure is meant to be propogoated when you register a domain staticsafe: somebody on [outages] reported issues with ultradns mercutio: to have mulitple name servers in different locations
ultradns use different locations for primary/secondary
they're at least smarter than "timewarner"
(although they may not in some locations for all i know)
ultradns has terrible routing from my country
i emailed them about it
didn't do any good
it's like 300 msec away or something
err to the primary
it's like half that to the secondary staticsafe: new zealand eh mercutio: yip
and they're routing to hong kong i think staticsafe: ultradns is heavily anycasted but not enough for users in nz i suppose mnathani: Anyone watch the debate? mercutio: static: but the anycasting gives worse routing!
that's the issue staticsafe: heh mercutio: i think it gave some packet loss too
what's their primary dns?
i'm assuming their primary domain may not use it staticsafe: udns1.ultradns.net and udns2.ultradns.net. mercutio: oh sweet it's fixed :)
primary is like 26 msec
secondary like 170
156.154.70.1 whatever it is is 145
i found that when googling but i think it's recursive
28 msec query time ***: jlgaddis has joined #arpnetworks
jlgaddis has quit IRC (Changing host)
jlgaddis has joined #arpnetworks
Ehtyar has quit IRC (Quit: Never look down on someone unless you're helping them up.)