up_the_irons: any chance you could provision the replacement VM in parallel with the old one still running (with temporary addresses, v6 would be sufficient)? the old one is still running fine, so I could restore my backup and then cut over once that's done
fIorz: should be doable.  just tell me which IPv6 I should assign.
up_the_irons: you mean from my subnet? just take a random one, chances are negligible that you'll hit one that's in use? ;-)
i suppose you're right ;)
OK, I have 1 VPS ahead of you, then I'll work on yours
sure, it's not particularly urgent as the old vm seems to be working just fine so far, and I'll be afk in a moment anyway ...
ok
another customer said a similar thing, that the VM was running fine.  But we did a ZFS snapshot of all VMs on that host and tried to transfer the data off, and OH MAN, was it corruption city.  'zpool -v status' reports your volume as affected.
well, it's mostly a tunnel endpoint, openvpn is probably in the pagecache, so chances are you could more or less remove the disk completely and it would still work fine ;-)
hahaha
fair enough
but, yeah, bad luck indeed with two disks failing at the same time ... same manufacturer/manufacturing batch?
it all came as a set.  a bunch of HP disks in a 25-bay server.  _really_ weird the failures were from the same raidz group.
makes me want to put a Ceph cluster out in Frankfurt
alas the initial server count needed is high
MIne just froze right now when tried to do ps aux :)
i'm not surprised
or it died even
yeppers. died hehe
oh well.. wanted to backup bgpd.conf
yeah timeout wait for disk kinda.. kernel outputs.
want me to reply to email I guess?
if you feel you're done with it and there's no more saving data, you can just tell me here
Well I cant save any data as it died :)
up_the_irons: but if you want you can put it to centos6.7?
or 6 :)
hangon.. I just got in again.
let me rsync what I need.
done :)
crap, I don't have the 6.x line in our templates in Frankfurt
I can work on that later today
if  you don't mind
ah ok. Np.
Ill shutdown bgp while its having related issues so i dont run into more diskusage perhaps.
ok
up_the_irons: my box is up again now. Thanks.
mrsaint: great!
no problem
up_the_irons: now syncing 5000 domains again :)
what do you do? :)
is that for like.. dns.. or?
think it's time for me to finally sleep...
g'nite all
up_the_irons: yeah dns and webanycasted service.
mrsaint: Anycast web stuff? How stable do you find the internet for that? Many problems with TCP sessions dying because they change routing?
plett: works just fine.. static content and varnish that do cache alot.
and I choose my providers for my servers with care. Specially take them with direct connects to either me or my providers..
there's actually a lot less varying with routes than there used to be.
and a lot of them come from brief outages where web page not loading right sucks anyway
and people reload
or go somewhere else
yeah.. And would be same problem anycasted or not.. Only times I would have connection problem would be when the node cant reach backend and it has not yet happend.. :)
yeah that seemse to be normal cloudflare issue
the node can't reach the backend
(cloudflare are anycasting tcp too)
yeah
reddit is the site i most have issues with
but their backend usually responds but not positively :)
my own site is dynamic but alot of it is static so works fine anyway. so we see. I only have gamesites and their png/gif content sofar.
but the dns works lovely really nice..
Yeah, anycast dns is a good solution. I'm not sold on doing it for web traffic though :)
i'm not sold on tcp for web traffic
with people expecting 2 second load times etc, slow resend etc delays really hurts
and it'd be better if the client knew about multiple places to source the same data from, and to have good web caches etc
mercutio: It's fine, just don't live somewhere a long RTT away from the web server ;)
linux decreasing the initial timeout from 3 seconds to 1 seconds really helps
but yeah high rtt connections really don't want packet loss
but then you end up with 1 second+ pings on 3g
because they don't want to drop packets
3G networks here are typically 200ms ping times. 4G are closer to 10 or 20ms
dc hsdpa or whatever is around 60 to 80 msec
but often there's some congestion
i found it went up when there was a power cut..
which is when i wanted to use it, and it was 1+ second.
i tried a few random times afterwards and wasn't nearly as bad
4g is here now at least.
plett: if you have a web you want to try you can try it for free if you wish. Prefer backend at a location where I have good connectivity to tho. To secure uptime.
aw, got momentarily excited to see that gandi has .nets for $2, but the promo doesn't apply to renewals.  :P
mrsaint: ah cool
up_the_irons: yeah works well.
nice
up_the_irons: re ceph: yeah, probably the best solution to that problem--but be careful with the crypto, it's terrible (unless they have redesigned that in the last year or so since when I did a rough review of the code)
we aren't using the crypto in ceph
that's probably the right thing to do :-)
:)