***: jcv_ has quit IRC (Quit: leaving)
jcv has joined #arpnetworks
_iwc has joined #arpnetworks
_iwc has quit IRC (K-Lined)
avj has joined #arpnetworks avj: anyone else by chance on kvr26? i initiated a reboot from within the VM, and not only did it not come back but i can't connect to the remote serial console or vnc over ssh (which i've used many times before)
i've tried a hard shutdown and power-on, but nothin'
i did submit a proper support request as well, just figured i'd pop in here too lyarick: avj: did you shutdown through the web interface or over the ssh console? avj: my initial reboot was initialized from a vnc-over-ssh session from the VM itself
and from there the vnc session crapped out, and i figured something got hosed in there
so i initiated a hard shutdown from the ssh console session lyarick: i've seen issues via the web interface.. console almost always works.
maybe try to boot again via the console?
AFAIK, hard shutdown will make it not respond to VNC avj: yeah, i tried that a few times. attempted initiating a boot, but not only does the VM never come up, i also cannot reconnect to the console in any way grody: just try to start it back and keep hitting up VNC - it should at least get to a 'post' stage on the vBIOS
not seen or had that issue myself
well, when i broke my kernel config - but was able to get in via VNC and fix avj: no dice, i've tried that several times already BryceBot: That's what she said!! grody: BryceBot, no BryceBot: Oh, okay... I'm sorry. 'no dice, i've tried that several times already' grody: that blows :( avj: it's as if the boot is failing
and of course, by that i mean the VM startup, not the OS boot grody: i think once, it took a little over a minute for any of the portal buttons to work once
but that was ages ago
but the console access yea..
KB just suggest power off if shutdown doesn't work, wait 60 seconds, power up, wait 60 seconds.. repeat if doesn't work
but it sounds like thats what you have been doing avj: certainly not at that frequency, but i've tried probably about 10-15 times over the last couple of hours
up_the_irons: ding ding up_the_irons: avj: I'll take a look at that shortly avj: hey, sweet. thank you.
up_the_irons: mercutio just grabbed my ticket from earlier. looks like it was a kvm hiccup. all is well now. thanks. up_the_irons: avj: yeah we worked on it a bit together. glad all's good now! avj: oh, nice. was this a fairly atypical failure mode or something more common? mercutio: atypical.
hmm php7 seems to be coming out in a month's time.
i am curious how quickly people will shift to it, it's meant to give around 2x speed improvement.
but historically a lot of people have kept using old php versions. brycec: I can count the number of PHP6 installs/users I know on my non-existent tail. mercutio: php6 was skipped brycec: good point mercutio: lots of people are using 5.3/5.4 though brycec: (Last time i was paying attention to such things, PHP 6 was still going to be a thing.) mercutio: and there's still 5.2
hmm www.php.net doesn't work brycec: works for me mercutio: invalid cert here
oh without https it works brycec: you didn't say you were trying https... why are you trying https? :p mercutio: yeh it's up to 5.6
everything is https now
i dunno brycec: btw it's secure.php.net mercutio: google links to https brycec: (I hit https://php.net and was bounced to secure.php.net)
...you googled php.net? lol mercutio: i think i googled php brycec: I hit https://www.php.net too and was bounced to secure.php.net mercutio: you're right, https://php.net redirects to secure.php.net
and works.
php.net and www.php.net give different ipv4 addresses
but they give the same ipv6 address. brycec: * Connected to www.php.net (2a02:cb41::7) port 443 (#0)
:P mercutio: heh
anyway, php 5.6 is current
http://w3techs.com/technologies/details/pl-php/5/all
and this suggests that php 5.3 is the most common version brycec: php.net has address 72.52.91.14
php.net has IPv6 address 2a02:cb41::7
both work fine for me, https validates etc mercutio: yip php.net has that address for me too
but www.php.net has address 202.78.240.39 brycec: www.php.net has address 72.52.91.14
www.php.net has address 64.71.164.5
www.php.net has address 216.194.115.245
www.php.net has address 208.69.120.58 mercutio: weird i only get one address for it
ahh if i query 4.2.2.2 i get what you get
if i query other nz servers i get that single address too, and it seems to go to nz
so yeh they're doing some weird geoip stuff mnathani_: what do folks think about https://www.reddit.com/r/sysadmin/comments/3ofth5/dell_buying_emc_for_67_billion/ RandalSchwartz: "dude... you're getting bought by a DELL!" mjp_: makes no difference.. brycec: "That's a lot of money. I'd let someone buy me for $67 billion."
(is what I think) mercutio: i think there needs to be more competition in such spaces.
dell was already selling emc stuff a lot though
http://www.theregister.co.uk/2015/10/12/linux_networking_api_showing_its_age/ BryceBot: Title: "BBC bypasses Linux kernel to make streaming videos flow • The Register" mercutio: showing it's age is a weird way to put it
considering linux's network stack is probably the most current in the world mjp_: hehe, theres probably a good reason netflix doesnt use it mercutio: they're using netmap
i think
i don't think netflix has any particular reason to use freebsd over linux mjp_: they probably just flipped a coin brycec: "This was selected for its balance of stability and features, a strong development community and staff expertise." Netflix on their choice of FreeBSD
https://lists.freebsd.org/pipermail/freebsd-stable/2012-June/068129.html
which can be interpreted as "We know FreeBSD, Linux's ecosystem is kinda shitty, and FreeBSD works very well." mjp_: im thinking its the TCP/IP stack. any *nix OS can run apache/nginx and mount in NFS storage mnathani_: why doesnt chrome wrap long lines as in this post: https://lists.freebsd.org/pipermail/freebsd-stable/2012-June/068215.html mjp_: <pre> mercutio: linux has a better tcp/ip stack than freebsd.
but stability wise freebsd is more conservative. mjp_: source? RandalSchwartz: yeah - I'd think any BSD would have a better TCP/IP stack than Linux
since it has the *original* code as its core.
as I recall, freebsd was also the original reference implementation for ipv6 mercutio: a lot of evolution has happende then
since then
i'm not sure of good online soruces
damn it's hard to google good information
some people say that freebsd has better tcp/ip stack because pf is better than iptables. mjp_: FreeBSD isnt good at a lot of things but the networking stack isnt one of them :) mercutio: hmm freebsd has multiple network issues
like memory pool fragmentation when using large mtu's.
for low/light load normal situations linux will usually give faster transfer rates than freebsd.
for heavy load better testing would be needed :) mjp_: source? mnathani_: how much does efficient device drivers for the NICS come into play? mercutio: i don't think freebsd is that bad network wise.
just get two hosts one with freebsd and one with linux, do some testing
and you'll find the linux host giving better speeds.
mnathani_: not much for gigabit. mjp_: i have done that exact testing, and I got FreeBSD to outperform mercutio: interesting.
did you use cubic congestion control with freebsd? mjp_: 2 exact hosts at OVH (canada) doing 100mb/1000mb file transfers to AU
net.inet.tcp.cc.algorithm=htcp mercutio: hmm
and what'd you use on linux?
did you try that new congestion control in freebsd btw? mjp_: these are all the options fyi: http://pastebin.com/mzizxY8q mercutio: i found it went worse than cubic mjp_: newreno? mercutio: nope
uhh
it was in like 9.2 or 9.3 mjp_: out of the box freebsd didnt perform as well, but with tuning it was able to outperform. around 20% faster peak transfer speed mercutio: https://www.freebsd.org/cgi/man.cgi?query=cc_chd mjp_: also faster ramp-up with HTCP mercutio: maybe this was it
did you tune linux too? mjp_: no, the linux box was a friends mercutio: ahh
well linux probably should have been tuned too :) mjp_: to be fair, probably :) mercutio: you're at like 160 msec latency?
so if you want 100 megabit then you need pretty large window sizes mjp_: the box was 1Gbps mercutio: in AU too? mjp_: i dont have it anymore as i didnt use it mercutio: hangon would be more than 160
dunno what i was thinking :)
more like 320.
so yeah you probably want at least 4megabyte window sizes
i think linux defaults to 2 megabyte
also ovh used to have bad peering
if it was before they expanded into canada
so there probably was loss mjp_: OVH is france isnt as good as canada, for AU people anyway mercutio: is there a test dl somewhere?
it looks to be about 80 megabit to proof.ovh.net
from new zealand
and about 150 megabit to arp
one provider here has bad routing to ovh though
people were complaining because it was going via asia mjp_: yeah its slow mercutio: yeh from what i've read about australian internet it seemed slow in general?
like most people get around 10 megabit? RandalSchwartz: it's slow to the rest of the world mercutio: https://en.wikipedia.org/wiki/List_of_countries_by_Internet_connection_speeds
weird BryceBot: List of countries by Internet connection speeds :: This list of countries by Internet connection speed lists the Akamai State of the Internet Report "Average Connection Speed" (ACS) of TCP-level Internet access by end-users accessing content on the Akamai content delivery network. According to Akamai, ACS is not indicative of overall broadband connection speed because it is mitigated by several factors such as parallel web browser requests, small mercutio: oh it's akamai
i was thinking those numbers looked funny mjp_: i have 100Mb/40Mb fibre at home and 10Gbps fibre at work, so not too slow :)
i can max out my fibre out home pulling data from the US without any problems
some areas have shitty speeds due to using ADSL on old phone lines or oversubscribed exchanges etc mercutio: yeah i heard that the exchange areas are huge there
oh and the oversubscription
then it's like $20/megabit or more for backhaul for fibre?
so probably will be bad too
err for nbn fibre
100/40 sounds good though
doesn't seem typical though mjp_: not sure about the backhaul costs
hopefully FTTP will be rolled out everywhere and FTTN will be canned
there is also quite a bit of fibre around. someone at work reckons it has been uncapped now and can get ~250Mbps down, faster than my fibre :/ mercutio: Each 200 megabit CVC will cost 200 x $20 = $4000 per month (at wholesale, plus GST) to connect (for an effective rate of almost half a million dollars per month for an RSP connected to all 121 points of interconnect) mjp_: is that still accurate? Thursday, July 21st, 2011 mercutio: i haven't heard anything to the contrary
there was a big spiel somewhere about the ont stuff etc
new zealand is doing similar fibre stuff to australia
but we seem to be doing it slightly less bad.
(i still think it coudl be better) mjp_: I pay $100/month for 1TB of data which i dont think is too bad for fibre mercutio: i use about 500gb/month i think
but i don't have cap
but yeah $100/month for 1tb data 100/40 seems ok
it was $109/month through iinet when i googled :)
i'm not sure what providers are good though
i just know whatever provider my friend had wasn't very good but was unlimited mjp_: looks like prices have risen to $119/month mercutio: oh i was looking at 500gb
do you use iinet? mjp_: yep mercutio: at least they peer
but ping seems on the high side mjp_: seems ok to me mercutio: it's 240 msec from japan
and 160 from arp
it looked like
i dunno, is there a good sydney ip to ping on their network? RandalSchwartz: ... http://www.cityofsydney.nsw.gov.au mercutio: that's not on iinet RandalSchwartz: ahh, no idea then mercutio: it hasn't got wonderful routing either it seems milki: nsfw.gov.au? mercutio: nsfw? :)
ahh bigpond.com.au
seems to be sydney, and have better routing ***: tabthorpe has quit IRC (Read error: Connection reset by peer)
tabthorpe has joined #arpnetworks
toeshred__ is now known as toeshred RandalSchwartz: yorkshire pudding day! ***: dj_goku has quit IRC (Read error: Connection reset by peer)
dj_goku has joined #arpnetworks