@e0qdk
I can only check federation with our mastodon server (pol.social), and now is very quickly (about seconds).
Queue on kbin.social is empty, so everything should work instant. Maybe lemmy.world add some limitation due to problem with kbin.social.
Feel free to add if you encounter some problems with infrastructure ;)
@grzesiek@rower miast miastem… mnie kiedyś zdziwiło że masa rowerzystów na DDR omijała rozwalone piwo. Nikt się nie zatrzymał. Dopiero byłem pierwszy… minuta żeby to przekopać w okolice trawnika…
@Burn1ngBull3t There are many transaction in Fedi, that why Mastodon and Kbin use SQL for standard operation. But in future maybe will be some place to something like Elasticsearch for full text search
@digitallyfree
Yes, but this is normal CPU for desktop not for servers that why we now waiting for sever class CPU. This one is working now on almost 100% load all day and night :) kbin.social have massive traffic, but we try to keep it alive ;)
@digitallyfree
now is core i9 with 64gb DDR5 and two samsung nvme (2TB each) on software raid.
but we just order a few machines with xeon-s ... so we try make some cluster that can grow. @haubles@vvuksan@renchap@ernest@huskola
K8S is one of option, but we decided to use some mix of bare-metal and docker swarm.
Almost everything is prepare to grow horizontally. Only (like always) problem is in database, and also we want to have flexible software that run on big cluster and small node without changes in code.
Give us few days, and after that we will show something ;)
@melroy we cache many things on varnish and CDN, and also we do many database optimization, but I think all I send to you.
And of corse many optimization in database configuration, disk optimization, PHP optimization. All best practice that I know I put there.
But if someone need help I can help with that.
@jwr1 @arotrios @jerry @vaguerant
@e0qdk
I think is Ernest instance.
We as @ftdl works only on Kbin.social
I will ask Ernest about Karab.in on some meeting.
@ernest @look997
@lwadmin hi, I will try to reach out Ernest to faster fix it.
Thanks for information!
@piotrsikora@lemmy.world test222
@piotrsikora@kbin.social
@Dave hi! I’am working with kbin.social infrastructure, and we now fixed some issues. Can you try to re-enable federation with kbin.social?
@e0qdk
I can only check federation with our mastodon server (pol.social), and now is very quickly (about seconds).
Queue on kbin.social is empty, so everything should work instant. Maybe lemmy.world add some limitation due to problem with kbin.social.
Feel free to add if you encounter some problems with infrastructure ;)
@ernest @rhythmisaprancer
@rhythmisaprancer we fixed some issues in infrastructure. Thanks for feedback :) @ernest
@grzesiek @rower miast miastem… mnie kiedyś zdziwiło że masa rowerzystów na DDR omijała rozwalone piwo. Nikt się nie zatrzymał. Dopiero byłem pierwszy… minuta żeby to przekopać w okolice trawnika…
P.S.
Szacun za akcje!
@Burn1ngBull3t There are many transaction in Fedi, that why Mastodon and Kbin use SQL for standard operation. But in future maybe will be some place to something like Elasticsearch for full text search
@EROLoLICON
Yeah… cost will be much higher than €100 every month ;)
Stay tuned for updates ;)
@haubles @vvuksan @renchap @ernest @huskola @digitallyfree
@digitallyfree
Yes, but this is normal CPU for desktop not for servers that why we now waiting for sever class CPU. This one is working now on almost 100% load all day and night :) kbin.social have massive traffic, but we try to keep it alive ;)
@haubles @vvuksan @renchap @ernest @huskola
@digitallyfree
now is core i9 with 64gb DDR5 and two samsung nvme (2TB each) on software raid.
but we just order a few machines with xeon-s ... so we try make some cluster that can grow.
@haubles @vvuksan @renchap @ernest @huskola
@digitallyfree
Yes, and we are going to change it in a few days:)
@haubles @vvuksan @renchap @ernest @huskola
Hi @Badabinski
K8S is one of option, but we decided to use some mix of bare-metal and docker swarm.
Almost everything is prepare to grow horizontally. Only (like always) problem is in database, and also we want to have flexible software that run on big cluster and small node without changes in code.
Give us few days, and after that we will show something ;)
@haubles @vvuksan @renchap @ernest