ceph is "reasonably good" for performance. that "reasonably good" performance continues as load increases so there doesn't tend to be bad performance but never extremely good. basically the latency increases slightly, but you can parallelise it.. so if you have a backup running say as you run your normal server load there'll be less impact. with writes it has to send the writes to 3 destinations and only says it's written when it's gnoe to all 3, so your performance is as good as the slowest disk/server. so instead of writing to 3 hard disks say then 2 parity disks it'll write 3 whole copies to 3 hard-disks and those 3 hard-disks are on 3 different servers across network anyone know anyone that wants to buy APC NetShelter cabinets? The ones we have in our current cage, we own them, but will not have any use for them after 4/30. Will sell for cheap if someone comes and picks them up! Who needs cabinets when you have ARP Networks! LOL The only place I could use a cab these days is across the US, sorry. oh, another question, are you supposed to be able to run VMs on top of ARP Thunder? since it is marketed as "dedicated" service yeah you can acf_ nested virtualisation Oh nice! Can we have some photos of the new cage? :D I thought it was 2019. Why does Facebook still display pages like 1280x1024 resolution when I have 1920x1080 ? Empty bars on both sides hav eyou got adblock? ususally they stick ads on the side yea, ublock origin try disabling it temp? there's a power button on the tray thing up top right normally https://imgur.com/a/WeDdyNR shows like a good 150-200 pixels wasted on either side of the middle content even with ublock origin disabled odd it's tiny too go to 150% zoom LoL now its covering more pixels but the content is HUGE heh makes it easy to read now you don't need a 50" tv for facebook Thought you had a largish monitor only 27" 4k even i don't use my 4k monitor I think mine are like 25" i use 1440p x2 the 4k monitor was only 28" and it had worse colours my tablet is 2560x1600 i hadn't used it in ages, just used it last night for sleep tracking, think i'll go back to that. but tablets are great for sleep tracking as they have big battery and you don't find your cellphone with hardly any charge in the morning can an ipad be used for sleep tracking? what would the app be called yeh probably i use sleep as android i suspect there'll be a sleep as ios or such (Obligatory: "It's 2019, why are you still using Facebook") iSleep? LoL brycec I just got accepted into a Sleep Apnea Group on Facebook what i like about sleep as android is you can set an alarm for say 8 hours and it'll wake you up up to half an hour early while you're in the light part of your sleep mercutio: Something odd happening with zeit.arpnetworks.com... It's suddenly getting inundated in traffic. I'm not sure what yet, just that it tripped its "over 4GB received in the last hour" alarm I have setup. brycec: erk it's not open somehow? err to amplification attacks i mean Shouldn't be :/ Trying to use Facebook to see what others have in terms of experience and advice what does tcpdump show? valid ntp queries / randomized source addresses? mercutio: It also seems to have stopped... mnathani: I didn't get a tcpdump in time It's back to its usual background hum of a few-hundred kb/s it shows unsurprisignly lots of ntp traffic Good thing you have netflow then :-) rx: 355 kbit/s 500 p/s tx: 351 kbit/s 495 p/s oh wow did you check out the tcpdump? I did not, why? it's got a tiny amount of ipv6 and the rest is ipv4 the ratio is crazy high towards ipv4 <-- this is my surprised face. i thought more people had ipv6 by now :) come on it's 2019 time to adopt ipv6! protip: Remember to exclude SSH traffic when tcpdumping hah been there done that I did an exercise a few weeks ago reaching out to local ISPs to see if they offered IPv6 in their Residential Service not one of them did yeah it's kindof amusing.. people like google and facebook have done ipv6 for ages and more and more people host with ipv6 support but the number of clients doing ipv6 is still pretty low and it's the clients that really need ipv6 more than the servers because you need to dedicate an ip to a server for ipv4 anyway but when clients can talk to each directly with an IP it's great! Anyways, seems back to normal... Didn't know if there had been some weird routing loop or other migration-related anomaly. there shouldn't have been any changes to it recently might be a buggy client or something One helluva buggy client to spam 5GB in less than an hour, but I can see it. (For reference, monitoring reported >3000pps and 10MB/s) s/3000/8000/ (For reference, monitoring reported >8000pps and 10MB/s) s/3000/8000/ | s/10/27/ (For reference, monitoring reported >8000pps and 27MB/s) bbl (Helps when I look at the max)