***: data has quit IRC (Quit: Leaving)
Guest95570 has quit IRC (Ping timeout: 264 seconds)
mnathani_ has quit IRC (Ping timeout: 250 seconds)
Guest95570 has joined #arpnetworks
mnathani_ has joined #arpnetworks
Guest95570 has quit IRC (Ping timeout: 256 seconds)
_iwc has joined #arpnetworks grody: it did spike at 93C
then dmesg was flooded with overheating, scaling down blah blah
then bham
seems to be holding itself together, thank goodness.. can't afford that loss just yet mercutio: weird ***: dj_goku has quit IRC (Read error: Connection reset by peer)
dj_goku_ has joined #arpnetworks -: grody must stop using bleeding edge code grody: last 3 builds of openwrt for my pocket router has been laden with an array of annoying bugs :( ***: brycec is now known as Big_Pimpin
Big_Pimpin is now known as brycec
neish_ has joined #arpnetworks
ameise has joined #arpnetworks
phlux has quit IRC (*.net *.split)
rendrag has quit IRC (*.net *.split)
ant has quit IRC (*.net *.split)
m0unds has quit IRC (*.net *.split)
NiTe has quit IRC (*.net *.split)
staticsafe has quit IRC (*.net *.split)
neish has quit IRC (*.net *.split)
daca has quit IRC (*.net *.split)
ameise is now known as ant
rendrag has joined #arpnetworks
phlux has joined #arpnetworks
m0unds has joined #arpnetworks
staticsafe has joined #arpnetworks
daca has joined #arpnetworks
NiTe has joined #arpnetworks
Guest95570 has joined #arpnetworks staticsafe: hrm v4 connectivity between my ARP Networks VM and my Nagios instance seems to be broken somewhere - http://paste.ee/p/1YMMA meingtsla: I'm hearing rumblings of Any2 connectivity issues since about 1:45pm PDT kellytk: I've not heard of that provider. Transit? meingtsla: No, peering exchange kellytk: That makes more sense mercutio: static: that looks like a networklayer issue
a forward problem from cs01.lax01.networklayer.com
unless linode hvae done something silly like have rpf issue as it's showing router-1-dal.linode.com and router-2-dal.linode.com
meingtsla: where did you hear that?
http://lg.arpnetworks.com/cgi-bin/bgplg?cmd=show+ip+bgp&req=45.33.23.57
so softlayer are linode? staticsafe: Linode is using Softlayer's network mercutio: i'd ticket linode given those two traces.
last i knew linode don't have looking glass.
looks like softlayer do though
erk their dallas node can reach both arp in general and your vm it seems brycec: I don't suppose http://apps.fs.usda.gov/ArcX/rest/info?f=json loads for any one else? (vs immediate connection reset or timeout) mercutio: but they don't give linkable urls with their lg. (http://lg.softlayer.cmo)
yeh not loading for me either brycec: thx mercutio: it actually just hangs
rather than connection reset brycec: I've been getting both mercutio: ahh
maybe their load balancer is misbehaving while web servers are unreachable
ie it's overloading because of so many connections trying to work
static: lax01 can't reach arp, lax02 can. i have no idea of what ip lax01 is on and reverse dns lookup from your traceroute doesn't have forward lookup
bbr01 vs bbr02, not lax01 vs lax02 staticsafe: hm mercutio: found it, the ip is 206.72.210.131
which is the any2ix ip address, which isn't advertised so isn't necessarily an issue
(because it's just used for connecting locally to other hosts, and forwarding traffic normally; they don't have an ip address just for traceroute, so it's using the facing address) meingtsla: mercutio: Email from CoreSite mercutio: meingtsla: ahh
i don't know who you are :)
but most people don't seem to know what any2ix is brycec: (Now I'm getting connection resets, woo) mercutio: yeah i wouldn't worry about how it manifests
it's likely to be the same core issue brycec: It's inconvenient is what it is :p mercutio: welcome to life :)
$136 usd is cheap for 480gb ssd brycec: Yes
Yes it is...
What form factor? mercutio: i'm thinking about going to 3 way raid-z with 3 of them
2.5" 7mm
http://www.amazon.com/gp/product/B00M8ABFX6?psc=1&redirect=true&ref_=ox_sc_act_title_1&smid=ATVPDKIKX0DER BryceBot: Amazon: "SanDisk Ultra II 480GB SATA III 2.5-Inch 7mm Height Solid State Drive (SSD) With Read Up To 550MB/s- SDSSDHII-480G-G25" mercutio: it's tlc, but i have two and they seem to go better than the samsungs
err better than the samsung 840 evos that were in there too
does over a gb/sec for two of them in raid 0 for part of the disk
and the reset is mixed with what were samsung 840 evos
err 250gb samsung evo's
it's just some non imortant stuff that's raid0'ed
but i've had to secure erase etc the samsung's
because performance of the whole set was going down..
Timing buffered disk reads: 3240 MB in 3.00 seconds = 1079.71 MB/sec
that's two of them
people are always like up to 550, up to 540... but usualyl i find samsungs are more like 510mb/sec
Timing buffered disk reads: 1470 MB in 3.00 seconds = 489.50 MB/sec
like that's a samsug ssd
Timing buffered disk reads: 1550 MB in 3.00 seconds = 516.50 MB/sec
and that's another samsung ssd
the first samsung was secure erased to get speed back up
it was going like 50mb/sec or something
(840 evo, the second is 850 evo) ***: dne has quit IRC (Ping timeout: 252 seconds)
dne has joined #arpnetworks m0unds: im hesitant to buy another samsung ssd ever again. had two of two fail and get replaced under warranty.
one stopped showing up in bios/uefi anymore and the other wigged out randomly
the hotel im staying at in durango co has a free cocktail hour. time for booze. mnathani: @weather yyz BryceBot: Toronto-Pearson International, Ontario: Mostly Cloudy ☁ 75°F (24°C), Humidity: 73%, Wind: From the Variable at 2 MPH -- For more details including the forecast and almanac, see http://www.wunderground.com/cgi-bin/findweather/getForecast?query=43.67722321,-79.63055420 or re-request this with: @weather -v yyz mercutio: yeh sandisk actually don't seem too bad.
they have worse random performance than samsung, but it's not like you're going to notice it with normal work loads.
i'm a bit hesitant about these cheap ssd plus ssd's with no reviews though grody: i have two ocz ssds and they seem fine
one is nearly two years and going strong
i dont see anywhere near those speeds though, mainly 'cause i run them on SATA 2 acf_: hm. I've had three Samsung SSDs for a couple of years, and they're all still fine
very light workload though
and no power failures m0unds: mercutio: yeah, i replaced my samsung drives with sandisk extreme 2s or something
acf_: my pc is on a ups and stuff. i guess maybe it was just a shitty batch or something mercutio: weird, i've only had ocz hard failure.
and i've got a few ssd's
this sandisk 480gb for $135 or something seems pretty sharp. m0unds: i have an intel 240 in my macbook and a corsair in the pc w the matched ones mercutio: wonder what amazon prime is going to bring though m0unds: keep dropping keystrokes. im connected to my box via lte on my phone mercutio: that seems strange, are you using that weird udp thing
or the lag is just disconcerting? m0unds: i think its just my keyboard
fleksy screws up mercutio: my error rate has gone down a bit from shifting back to blue from red.
err that's cherry mx. m0unds: this is on my phone
so touchscreen kb
juicessh
is the ssh app im usinf
using mnathani_: is there a good way to remember which argument goes first when executing a "ln -s path1 path2" command ? I frequently enter it incorrectly mercutio: not that i know of :)
i've screwed that up before too
maybe think think of -s as source
although it's not
so that it's ln -s source destination
just trying to think of something rememberable
although it depends how you think of source/destination
ie, source you could consider to be the original file, or the symlink that points to the original file
i cosndier it the original file, but it wouldn't be wrong to see it the other way around. grody: when you see 'file already exists' you know it's the wrong way :)
i almost always do it the wrong way, especially during file managemtn in the cli
got into habbit of link "this right here" "into this location here" now mnathani_: mercutio: thanks mercutio: grody: it doesn't always do that. grody: not always, especially when batching mercutio: oh, what i've hit is when it's a directory
so if if you do "mkdir somedir; ln -s notdir somedir" then you'll end up with a symlink of notdir to notdir in the symdir directory. mjp_: just remember its the opposite of whatever you think it is mercutio: haha mjp_: ln -s <real_thing> <link_that_points_to_it> mercutio: i'm trying to think up words that begin with s to link to "real thing"
static is my best one so far.
so static file, and dynamic link to it
probably better than source/destination
actually just thinking of it as ln -s <destination> <source> works better for me
it's always the little things that are complicated :) mnathani_: how about ln -s <existing path> <path or file to be created and linked to existing from first argument>
althought I might have them in the reverse order mercutio: not having whoisguard on domain attracts so much spam mnathani_: depends how popular the domain is I guess mercutio: notinthe.us mnathani_: saying the domain in here wouldnt really help as [FBI] would log it mercutio: haha idk it's already getting spam
it's targeted spam, so i don't know how much diff it'll make
there's probably some list of new domain names somewhere mnathani_: who is the registrar? mercutio: namecheap mnathani_: enom or enom reseller
ahh mercutio: my whoisguard domains don't get spam
but .us can't get whoisguard on. also strictly speaking i'm not meant to have a .us domain unless i'm in the US mnathani_: says so right in the domain
not in the us
lol mercutio: heh
it's been fine so far mnathani_: what do you think of softlayer as a dedicated provider? mercutio: well static just had issues with softlayer
like 7 hours ago
but yeah i dunno what they're like... i think they're big in dallas?
why not just use arp for dedicated servers? mnathani_: his issue was more with linode
a customer of softlayer I beleive mercutio: yeh could just be linode
you have a good point there
because the dallas softlayer looking glass worked. mnathani_: I would go with ARP, except my customer is looking for a Toronto dedicated server mercutio: ahh
toronto is another kettle of fish
i don't really know much about the area... mnathani_: and I like a provider that thas decent SLA, hardware replacement guarantees etc mercutio: but level of service could be quite different there (for better or for worse) acf_: I've been looking for Canada things recently mercutio: doesn't even ovh have that acf_: ovh has one location in Quebec mnathani_: amanah.com
those guys seem decent. Voip.ms use them mercutio: they were top link
i was put off by it saying 100tb
peer1 came to mind as a sensible idea
without too much thought mnathani_: softlayer only offer 500 gig on their base dedicated acf_: most of the providers I've seen are peer1 resellers
or ovh resellers mercutio: well peer1 are old
s/old/established/ BryceBot: <mercutio> well peer1 are established mercutio: yeh i wouldn't go with ovh acf_: what I was noticing
tracerouting places
most routes in Canada come back through US anyway
Vancouver -> Vancouver goes back through Seattle a lot mercutio: i suspect peer1 are expensive direct. mnathani_: 100 TB per month is like 304 mbit sustained for the whole month acf_: even Quebec -> Vancouver sometimes goes back through Seattle BryceBot: That's what she said!! mercutio: mnathani_: yeh, it means "oversold"
acf: that kind of stuff really bugs me.
in the US it's pretty common to take strange paths to some states.
like kansas is pretty basd.
bad
texas is pretty good generally. acf_: I was also looking at Centurylink cloud
because they have a PoP in Vancouver
(Savvis) mercutio: california it's mostly just san jose or los angeles leading to some weirdness, but it's not terrible.
ie it stays in state at least.
aren't savvis like cogent? acf_: Savvis always finds some way to route you around the other side of the country
ie, if you're tracing LAX -> LAX it will send you to NY mnathani_: Should I be looking at purchasing a dedicated and colocating somewhere? mercutio: hahaha
he.net used to be like that
mnathani_: yes.
well maybe ***: Guest95570 has quit IRC (Ping timeout: 240 seconds) mercutio: s/looking/considering/ BryceBot: <mnathani_> Should I be considering at purchasing a dedicated and colocating somewhere? acf_: I think HE had some of the best routing in Canada (from my traceroutes) mnathani_: I always get nervous when I am the one responsible to replace equipment acf_: Peer1 was OK too mnathani_: I guess remote hands to exist BryceBot: That's what she said!! mnathani_: BryceBot: no BryceBot: Oh, okay... I'm sorry. 'I guess remote hands to exist' mercutio: yeh he.net is terrible to asia, and used to have some weird stuff in the US, but i think it's not too bad in canada
it's around atlanta that it was most iffy iirc
and weird things like sj -> ny -> fl or something
mnathani_: just get 3 hosts
and host 2
then you have redundancy and a spare host if run into issues
l5520 type hosts are cheap on ebay
and good enough for non cpu demanding things
http://www.ebay.com/itm/HP-PROLIANT-DL320-G6-SERVER-505768-B21-XEON-SIX-CORE-X5650-2-67GHz-DVD-MULTI-/331584098520?pt=LH_DefaultDomain_0&hash=item4d33f3d0d8
what about something like that?
there are lots of options, and shipping varies a lot...
but if you don't need a lot of cpu the hp g6s all have ilo built in for remote serial / video mnathani_: bring your own ram mercutio: you need ddr3 ecc
which is about $100 US per 16gb i think
with single cpu you can run 6 or 9 slots depending if you care about maxing memory speed
with dual cpu 12 or 18
but you only really want to go dual cpu if virtualising or knowing you're running really scalable stuff mnathani_: something like : http://www.ebay.com/itm/16GB-4X4GB-DDR3-MEMORY-RAM-PC3-10600-ECC-REG-DIMM-/350444499824?pt=LH_DefaultDomain_0&hash=item51981eb370 mercutio: i'd shift to a e3 or e5 if cpu demands were high
yeh except if you want to mix with existing ram you have to check if it's registered or unregistered
if replacing all ram then it goes in sets of 3
so i'd do 3x8gb..
ie it's triple channel
http://www.ebay.com/itm/48GB-6X8GB-Hynix-HMT31GR7CFR4A-H9-2Rx4-PC3L-10600R-DDR3-1333-1-35V-/181768655729?pt=LH_DefaultDomain_0&hash=item2a52419371
you could go for something like that...
if you're not paying for power you may want to just go dual cpu, second hand it makes not much difference to price
then there's whether you want to go for 2.5" or 3.5" hard-drives
2.5" is handy for ssd's
but then you want to put a different raid controller in
http://www.ebay.com/itm/HP-Proliant-DL360-G6-Server-2-Quad-Core-X5570-2-93GHz-48GB-1-146HDD-/271928680850?pt=LH_DefaultDomain_0&hash=item3f50366192
that's already got 48gb of ram. mnathani_: those are way more perfomance than we need mercutio: and faster cpu for mhz, but quad instead of hex core. mnathani_: currently running on ; Intel(R) Xeon(R) CPU E31240 @ 3.30GHz mercutio: e3 is fast
the 5500/5600 are like half the speed of e3
dual cpu on the high end is similar in performance
as long as threadable mnathani_: 90% idle through the day though mercutio: i don't think e3s are cheap to buy second hand or new though
yeh if you don't need cpu you don't need cpu
my personal dedicated server is an i3 with 8gb of ram
it never seems slwo ***: Guest95570 has joined #arpnetworks mnathani_: any virtualization? mercutio: it's dl180g7 i think
nope
err dl120
http://www.ebay.com/itm/NEW-HP-Proliant-DL120G7-12Gb-250Gb-Xeon-3-1Ghz-1U-Rack-Mount-Server-/271929110697?pt=LH_DefaultDomain_0&hash=item3f503cf0a9
like that but i3
oh hmm that's an e3 for ok cost
main caveat is single psu
but they take 4x3.5" hard-disks
and i stuck a ssd in hiding behind the hard-drives for ssd cache
they don't do hw raid, but you get the normal intel sata ahci ports (no sas)
and although there are 4 bays, there are 6 ports.
also they don't take registered ram
so yeah just figure out requirements, and you can probably find something cheap mnathani_: whats decent in terms of price for 1U / 2U colocation? mercutio: completely depends on location
i'd ballpark $50 to $200 for 1u mnathani_: I would probably have to figure out commit rate and 95th percentile mercutio: yeah or just get gb
it's usually cheaper where electricity/real estate is cheaper
and more expensive when known brand is on it
i don't think toronto is a "cheap" city acf_: gb? mercutio: gigabytes. acf_: is 1U the 'cheapest' unit? mercutio: yeah
1u is what i linked
racks are usually 46 to 56u tall
err it seems some are 42
you may have power limit as well acf_: is colocation or renting a dedicated server usually cheaper? mercutio: colocation long term dedicated short term
but colocation is a bitch if you're not in the same city
or if you don't want to have to go there :) acf_: it's annoying you have to email for quotes for colo mercutio: tehre's also the thing of how many servers you want... if you want 1 or 2 servers then it's harder to have spares then if you want 7 or 8 servers. mnathani_: dedicated has the advantage that replacing parts is someone else's responsibility mercutio: arp is listing colo price on their web site
mnathani_: yeh
hah ok it says contact for quote
but yeah, it /was/ a lot cheaper than dedicated mnathani_: main page says $99 / month acf_: so do you usually ask the datacenter for a quote? mercutio: but now dedicated doesn't cost that much more through some peopl
you don't have to deal with othe rpeople going into the dc and access and so on acf_: and a transit provider? mercutio: well like somewhere like arp you get bundled transit
but some places are just data centres and you can choose your transit
but if you're getting 1u you don't want to buy separate transit
if you're getting a rack it makes sense though mnathani_: other datacenters you can get a crossconnect and buy transit directly from the providers mercutio: mnathani_: yeh but crossconnects can bel ike $200
plus the bandwidth cost acf_: $200 per month? mercutio: yeh acf_: woah mnathani_: like he said you would only do it if you had a rack or specific needs mercutio: yeh but if you pay $1000/month for transit then $200/month for cross connect it doesn't seem so bad.
it can vary a lot
but like where arp's located the data centre is spread across multiple buildings.
and like getting your own fibre between buildings is way more expensive than that acf_: so how is transit typically billed? mnathani_: commit rate and 95th percentile acf_: so commit rate is the minimum rate you have to pay for, regardless of what you actually use? mnathani_: I heard if you want gigabit, you need minimum 100 meg commit
right mercutio: acf: 95th or flat rate.
mnathani_: varies
i'm sure arp would do gigabit line with lower than 100 megabit commit
link acf_: what's a typical rate ($/mbit/s)? mercutio: acf: there isn't one
it varies
cogent and he.net can be as low as 50c/megabit
but that's on high commit levels
and in areas they have excess capacity
generally speaking if you want to get 4x10gigabit links you can get good pricing easily.
but if you want to get 100 megabit, don't even think about going direct. brycec: fwiw here's a cheap colo deal https://www.fdcservers.net/colocation.php
(they are *cheap* and not known for being "great") mercutio: level3 i've heard numbers like $3/megabit, but don't know anythign current acf_: do people like HE sell colocation outside of the datacenters they own? brycec: You can't sell someone else's space mercutio: dunno
fremont he.net has a really bad rep
brycec: you can resell brycec: (Well, you can resell technically, but that means you're selling your "own" generally) mercutio: so he.net might provide bandwidth and rack space in a dc they don't own
i wouldn't go with he.net primary for anything important acf_: in Canada they're looking pretty good mercutio: a blend is still probably better
http://www.caneris.com/Hosting#colocation
$125/month for 1u colocation,
wtf
10mbps port mnathani_: I have 10 meg upload at my residence
60 meg down
thats a joke mercutio: yeah it looks old
i'm sure they'd do 100 megabit for the same cost :/
i have no idea if they're good, it's just the first quote i found in toronto mnathani_: there is 100 megabit then there is 100 megabit you can actually use and burst to mercutio: i wouldn't worry about 100 megbit generally these days
i'd be more concerned with transit providers / peering etc.
oh amanah might be ok
it's not 100tb on all servers
but ouch $200
@exch 200 cad to usd BryceBot: 200 CAD -> 157.03579238298 USD (as of Tue, 14 Jul 2015 21:01:01 -0700) mercutio: oh not so bad
they charge $100 for 1u with 30tb/100 megabit
1 amp
what voltage is canada?
ahh 110 mnathani_: 110 v for res mercutio: yeh i didn't know if it was the same as US or not
so yeah 1 amp is not much power
it's at 1 yonge street
do you know where that is? BryceBot: That's what she said!! mnathani_: yea
its downtown mercutio: convenient? mnathani_: parking can be expensive
but yea mercutio: i wonder if they provide parking
well it's like $80 usd/month cheaper for colo mnathani_: https://www.amanah.com/Network/ mercutio: and you get 30tb instead of 10tb but you get 100megabit instead of gigabit
it's bell.ca from here
oh front street west is the main building i think mnathani_: Because we’re located so close to the American border, we’re able to offer both Canadian- and US- geolocated IP addresses
that is also downtown mercutio: that's not /because/ of that. mnathani_: convenient to get to also mercutio: swewet
well maybe go with colo there?
they're not on he.net
it doesn't look like
but he.net peer anyway :/
erk they're on cogent :/
y'know i don't know if they give you a vpn for ipmi BryceBot: <mercutio> ert they're f cfge t :/ acf_: BryceBot: ? mercutio: i think it's the y'know
y'know
hmm
erk
ok i'm lost what is it :)
cogent acf_: y'know i don't know if they give you a vpn for ipmi BryceBot: <mercutio> cfge t acf_: y'know i don't know if they give you a vpn for ipmi BryceBot: <mercutio> cfge t acf_: umm mercutio: i think he's laggy too acf_: y'know i don't BryceBot: <mercutio> i thint he's laggy too acf_: y'know i don brycec: Because y// is a valid perl operator
equivalent of tr// mercutio: oh it's y/
y'test/ brycec: y/t/r/ BryceBot: <mercutio> y'resr/ brycec: (and ' is a valid delimiter in sed)
(and BryceBot just mashes it all up together) mercutio: heh brycec: So above it was the two y'know i don't
that was y'' acf_: mercutio: what specifically is so bad about HE?
their routing has always seemed sane to me mercutio: acf_: they're a cheap non-premium provider with reduced redundancy acf_: ah ok mercutio: i don't think being with any single provider is good fwiw
it depends what expectations you have
like he.net links to EU were unusable when new york had that flooding.
whether that matters or not is the question
i mean it was big news about the flooding etc. it's not like it's all the time etc. acf_: yea.. for my purposes redundancy isn't all too important mercutio: their fremont data centre had a lot of stability issues too.
then he.net is probably fine acf_: it seems like HE buys a lot of transit from others mercutio: also he.net is more likely to hit issues with at&t, verizon etc. acf_: which makes them not as susceptible to the peering snafus mercutio: you reckon?
it's hard to know really acf_: I may be very wrong :P mercutio: ntt had some peering congestion acf_: with HE? mercutio: and they're meant to be tier1
nah
with at&t i think it was
i can't recall for sure acf_: Verizon
I was the one always complaining about that :P mercutio: oh
ok verizon :)
verizon and at&t seem the same to me :)
i know they're totally different
well different companies
similar attitudes :) acf_: Verizon is especially bad I think mercutio: so is at&t
cogent<->at&t congestion was shocking acf_: when two stubborn assholes peer..
Cogent still won't IPv6 peer with HE mercutio: why! acf_: you can't traceroute cogent.net from HE .. mercutio: what is bell.ca like? acf_: idk.. I live in the US for another month
then I'll find out :P mercutio: heh
tcore4-toronto21_hundredgige1-3-0-0.net.bell.ca
i've never seen anyone else say in their reverse lookups that they're using 100ge
he.net say 40gbe in some stuff i think acf_: http://paste.unixcube.org/k/b513bb
so it looks like HE buys transit from Verizon mnathani_: my route used to be he.net acf_: (or peers, but that seems very unlikely) mnathani_: but switched to level3 now
from Toronto to ARP acf_: so I'd hope they'd just pay up if they needed to upgrade the link mercutio: mnathani_: better or worse?
i don't have any routing issues to arp normally