the birth of flaqualakafluke.net is 2014-07-09 used rsync later upgrated to git 1march 2016 @2017-12-26 flaqualakafluke is alone i realised that because its git ssh and bash based, it will work on git bash i have to secure flaqualakafluke by at least replicating it once termux one vm @2017-07-06 salvaged added dlna renderer play nicely with voice not perfect but really cool wonder how can i combine these with MPD need to reactivate a VM on blackpearl.net now grendizer.local crazy idea about a zfs pool made of usb key. even usb2.0 could cap wifi N for a transportable fan less server rpi zero W otg usb 3.0 powered hub (several keys) zfs pool @2017-02-03 broken my hand, painfull writing. separated. now in a room without flaqualakafluke.net i use it remotely with my chromebook need a better strategy. at least need a pc. @2016-11-07 now with mosh as a good ssh alternative for flaqualakafluke i upgrated flatalatapouet to xiwi because it is more convenience that way. with mosh , and increased storage, i think i can afford to start RL/IF gaming on this network. not that the rpi can handle dwarfforteress. but its clone may. but in reality, i think it make so sense to versionise videogame because even ascii one have binary artifact. at least, dont use the same repository. need analysis. how to handle remote ascii gaming and keep savegame? ex: RL - IF - ZZT while they have binary, only the savegame will make increase the size over time and this wasnt a problem in the floppy disk time so it should be ok. then having a repo with all my game with dependency script and startup ready to mosh. seam that i need a second flaqualakafluke-like network @2016-11-01 chromebook is full. should upgrade flatalatapouet to xwi. and dispose of banana and the other vm. I can build a game one that comme with a game. new cell phone , should be rootable, new flaqualakafluke node in progress using chroot. So the phone may have local flaqualakafluke. flaqualakafluke should have an app install script. so all node have at least a basic set of tools. flaqualakafluke could comme with textbased games. like ZZT on dosemu. Some IF and Some RL. This machine was intended for coding textgame on the bus. dwarfsteroid - RL were you are piloting an asteroid through space, dig other and upgrade yours. blafurhtaggen - IF tts were you protect a city from invader that come from a train. you also send train to other city. YmcFace - IF without any common sense. findSara - RL zombi apocalyspe, find sara before she is killed in a village. logicly they will be coded in python using sqlite but... making it in haskell/erlang is a charming idea. With erlang and ncurses, i'll have async realtime network game. @2016-08-17 added on usb 3.0 32G stick and it work, as ext4 i plan to add a second one in a raid-like fuse system and i will simply move and symlink /home/majestic_nicolas_cage in it. @2016-07-28 still need upgrade 64$ 256gb usb https://www.amazon.ca/Lexar-JumpDrive-256GB-Flash-Drive/dp/B00S5V5PTC?ie=UTF8&creative=395285&creativeASIN=B00S5V5PTC&linkCode=df0&ref_=asc_df_B00S5V5PTC4418438&tag=shopbotca-20 20$ 32gb sd + adapter https://www.amazon.ca/Sandisk-Microsdhc-Speed-Microsd-Adapter/dp/B004G7D0EG?ie=UTF8&creative=395285&creativeASIN=B004G7D0EG&linkCode=df0&ref_=asc_df_B004G7D0EG4418438&tag=shopbotca-20 rpi 1sd 4usb hub? = 276$ @2016-06-20 lot of water under the bridge, had a journey through an unwanted limerence i am now modifying my VM architecture at werk and have some idea for a better flaqualakafluke btw flaqualakafluke was used as psychology tool in the journey without upgrade, current flaqualakafluke net can welcome three new nodes and many wifi nodes. (ddwrt) flaqualakafluke can welcome bigger sd card and usb storage ~15$ per 64GB stick, 4*64bg => 256GB MHDDFS there are 85$ca x86 fanless board that can become server maybe (150$) for the ram ssd and psu. combined with a powerful rpi(2-3). @2016-05-16 complete success now flaqualakafluke net have 3 slaves on 3 different context. @2106-05-15 banana is a chroot inside my chromebook. it have recovery script it turn out i cant prevent my chromebook to wipe everything but once in devmode i have ssh and scp and i can encrypt a chroot then i can store a flaqualakafluke node inside flaqualakafluke that can be restored in a cmd. its name will be... after a backup/restore (somewhat successfull). I am now considered online. It is name is flatalatapouet btw and banana to is integrated, scp is way more convenient than googledrive @2015-05-08 unexpected tax return bought 5 chromebooks and a new router (gbit over the wifi) this left the ddwrt free for a crazy project the old router will be part of the flaqualakafluke.net http://getchip.com/ is a potential contender for a cluster node 1ghz arm processor like android and iphone 512mb ram , 4gb storage , wifi and blutooth usb port when online a node may attempt to connect flaqualakafluke one reverse ssh connection that also sshfs to self then flaqualakafluke may use mhddfs to merge all node storage into one big note : mhddfs man warn against combining sshfs together @later from below load save and wazzup can update a file for a system health and wazzup can interpret it all local vm can safely open reverse ssh to flaqualakafluke @2016-05-02 now i can tunnel any port through ssh now i can reverse ssh if i want now i can run x11 through ssh flaqualakafluke.net is still operational and used blafuhrtaggen.net is a pain in the ass, will become a backup will add key to flaqualakafluke and a cron job for blafurhtaggen youbumtoo.net is still operational madafaka.net does not exist yet will be added alongside blafurhtaggen will be ubuntu lts 32bits with 3gb ram and 3 core wiil be used while on lan i wont make an instance in the boner.net because i cant manage its security @2016-04-06 not a bad idea but I thing docker can abstract network, then a central file server can be dockerised and shared in the network every container would have its own mac address and its own hostname in the dns. werever it boot it is available under hostname automounting concept for rpi swarm every hour all service shutdown for 15 minutes every nfs is unmounted every export is unexported every usb is unmounted then light indicate you can safely remove usb drive after 15minutes scan every usb mount what is found under its label name export it mount every export from host list then restart service @2016-03-15 the creation of couchMQ make couchdb a requirement once this accepted, it is important to globalise some tool located in node/peep/js/tool most of them are must cleandb savedb loaddb plog they must become available as cmdline and maybe ported to reveral language (at least python) cleandb must also be ported to couchMQ so it can trigger cleanup 2016-03-11##################################### flaqualakafluke.net has been migrated to git -fossil had issue -git worked out the box (had problem with false submodule, but fixed) -the phone is still not in the pool I now need a way to communicate between node, like messaging system couchdb is a rest DB, it work with any PL and is replicable (the way of flaqualakafluke.net) -need to expose couchdb to the network 2016-02-11##################################### refactor flaqualakafluke.net -learn to use fossil over ssh -make fossil server in flaqualakafluke -rewrite push and pull for use with fossil instead of rsync -update all node -add phone to the pool with cron job backup 2016-02-04##################################### using fossil over ssh keep data isolated from the infrastructure it is distributed, and will keep all my node in sync and error free the entire system will be contained in a file 2016-02-01##################################### rsync based synchronisation is risky -i forget to push before leaving -i could mix push and pull and loose my change in many possible way -i could forget to pull and loose change on server or on replicant using git though ssh make sure to git ignore artifact 2016-01-22##################################### it take 16 rpi2 to meet i7 they are only 80w hungry for whole system compute node sharing san model cant be efficient because of limited network bandwidth usb bandwidth is up to 200mbps if every 10minutes, all vm suspend all use space are backuped all usb unmount , let admin hotswap if needed wait for 5 minutes scan for usb to mount, can usb fs for dir struct they resume all vm contained in these, till next shutdown how: (cron job, docker load, docker save, rsync) cron job that loop for all /dev/sd[A-Z] mount it if exist scan for a particular dir struct > docker load $USB/$VM/img.docker -v /var:$USB/$VM/var cron job that stop all vm backup volume dir to 7zip and rotate it sync umount usb this will give 40minutes of work per hour, 20 minutes of hotswap downtime high resilence with abusive backup strategy smart network discovery can be achieved in static way, ip json file dont forged blinken light with ledatrix foreach node 2016-01-12##################################### replication is the key if data is recuperable, everything is alright 32gig is more than enough while non-stop operation is not required no need for a twin all machine keep its own data all machines backup backup strategy: (basicely rsync to a timestamp named dir) level1 internal usb storage mount sda0 check free space while not enough: erase earlyest backup dir backup level2 hotswapable usb storage from sdb0 to sdf0: mount, check free space, if enough free space: rsync umount break umount BOM: 32gig sd ~15$ 32gig usb * X+1 ~15$ usb 2.0 hub 2016-01-11##################################### requirement: reliability accessibility (A-P) from CAP theorem blinkenlight can afford: downtime #backup strategy slowliness cant afford: dataloss #backup strategy HW: robox cube ddwrt router rpi b+ 'flaqualakafluke' sd 8gb WISH: add blinkenlight to gpio 'flaqualakafluke' add rpi 2 b+ 'quad core 1gb ram' usb hub with usb stick as backup storage backup: once in while, usb stick will keep backup snapshot two usb stick will contain both a copy replication is the safe-guard about usb stick on rpi: they are mapped in order /dev/sda1 /dev/sdb1... http://elinux.org/RPi_Adding_USB_Drives#USB_Drive_Hierarchy hotplug: if the backup script mount/rsync/unmount then hot swap usb drive without reboot the virtually undestructible nature of usb stick network: once ddwrt emmanciped from desktop 4 lan port, 1 wan, +wifi wan long cable lan1 rpi lan2 rpi2 lan3 free lan4 usb lan adapter for management wifi broadcast for intranet of course, port redirection BOM: rpi2 ~60$ amazon w/case or ac sd32g ~15$ usb32g ~15$ ############################################### ############################################### HW: raspberry pi 2 B have quad-core cortex A7 at 900mhz -> 6.7DGIPS/10watts performance 1.9DMIPS per mhz -> 4*1.7DGIPS it have 1gb ram 10watts max banana pi m3 have octa-core arm cortex a7 at 2ghz -> 30DGIPS/10watts performance 1.9DMIPS per mhz -> 8*3.8DGIPS it have 2gb ram it have sata and gbit eth 10watts max while i7 have quad core hyperthread at 3ghz -> 84DGIPS/150watts performance 7DMIPS per mhz -> 8*21DGIPS >150watts banana pi r1 is a router with dual core a7 , sata, 1gig ram and 5 gigabit eth port. design of a toy supercomputer: bpi-r1 8gb micro-sd 64gig ssd 4 * bpi-m3 4 * 8gb micro-sd 180mm fan sw : debian+docker cfg : ssd is shared using nfs over gbit est-perf : 80mb/sec volume per node "all node have its own channel, and ssd out perform the network" 30gips per node -> 120GIPS, like an i7 50watts powerdraw 500$ price, hardly build an i7 for this price usage: the router will randomly generate intermittence in the network. temporary disconnect node from its peer without removing the nfs, and sometime remove the nfs, maybe even lower the speed of a node if needed the router will also redirect some ip:port to another ip:port the nodes will run container ,mainly python/nodejs/erlang and little bit of C. the router will also coordinate all running container the router will stop all container and backup its harddrive then restart the container It is a network lab for distributed computin bom 2 : 8 port gigabit switch 35$ 5 port gigabit switch 25$ one banana hold the ssd one banana route EXPECTED PERFORMANCE it will be less powerfull than an atom per core so all process will need to be lightweight with m3 you have 30gips for less than 10watts at 100$ with rpi2 you have 7gips for less than 5watts at 55$ with i7 you have 120gips for less than 150watts at 700$ 16 rpi would perform like i7 with less than 80watts and would cost 880$ 4 m3 would cost 400$ and consune less than 40watts for the same. 1gbits for interprocess communication is slow. it was the speed of pci bus. enough to power a videocard, a 3dfx, sound card, network, storage at the same time. until 2005. it will be a bottleneck but not an unmanageable one. a tera operation can be reached with 35 m3. 350wars and 3500$