Notices by walfie (walfie@kirakiratter.com), page 3

@MasterDalK ?

I feel better because now the entire stack is version controlled and arbitrarily scalable across nodes instead of being on a snowflake server. and I actually have centralized logging and good monitoring instead of relying solely on `htop`.

if these nodes go down, or even the whole cluster, I can easily just `kubectl apply` all the container specs, the letsencrypt provisioning config, the nginx ingress rules, backup scripts, etc on a new cluster and restore postgres from backup. good times

@prototype27 now it is truly "cloud native"

??

仕事で学んだことは有用でした

???

新しいサーバーです！
以前は1つの大きなサーバーでしたけど今は3つの以前よりも小さいサーバー

cpu usage of the database node is 100% during the backup, but performance doesn't seem bad. also thankfully I get to debug at convenient times while everyone's asleep

おほよ

now to see if the database backup cron job crushes the server in 5 minutes

カツカツ

kirakiratter is now running on digitalocean kubernetes across 3 nodes, instead of a single linode box

メンテナンス終了(？)と思います

テスト

メンテナンスします

wow I finally used `mkfifo` for something. A kubernetes pod with 3 containers with a shared emptyDir volume:
1. busybox init container that does `mkfifo /tmp/pipe` and exits
2. postgres container that does pg_dump on a database and pipes it into /tmp/pipe
3. rclone container that pipes /tmp/pipe into `rclone rcat` to upload the database dump to remote storage

for some reason all my Spotify ads have British narrators now, for the past few days

anyway I've decided with this setup it's probably better to have 3 nodes with 1CPU + 2GB RAM each than it is to have 2 nodes with 2CPU + 2GB RAM each

kubernetes pod monitoring dashboard looking good!!!!