Post AUbvJCrJ2FCCdTXIBc by ro@raru.re
 (DIR) More posts by ro@raru.re
 (DIR) Post #AUbpImonFnVo9CmFZg by ro@raru.re
       2023-04-13T12:10:17Z
       
       1 likes, 0 repeats
       
       High-availability SQLite
       
 (DIR) Post #AUbpK8HMwqgezrmk8e by wizard@xyzzy.link
       2023-04-13T12:25:10.514219Z
       
       0 likes, 0 repeats
       
       @ro web scale
       
 (DIR) Post #AUbqezV8sZ1HoAjmIi by ro@raru.re
       2023-04-13T12:31:32Z
       
       1 likes, 0 repeats
       
       lol at their commitment> Is there a commitment to long term releases?> The v1 series will be maintained, improved and bug-fixed for the foreseeable future and backward compatibility is guaranteed.then one day they'll be like, oops didn't foresee this far!
       
 (DIR) Post #AUbqf1FGOZVTDVK7pQ by ro@raru.re
       2023-04-13T12:33:37Z
       
       0 likes, 0 repeats
       
       i'd actually consider using this if i had a task that requires it though
       
 (DIR) Post #AUbqgXwmIQAQvgPChk by wizard@xyzzy.link
       2023-04-13T12:40:26.616973Z
       
       0 likes, 0 repeats
       
       @ro wait this is a real thing?
       
 (DIR) Post #AUbqs4HiPDA4PNh0T2 by wizard@xyzzy.link
       2023-04-13T12:42:28.859690Z
       
       0 likes, 0 repeats
       
       @ro canonical loltheir use case is probably for distributing k8s cluster key-values, which maybe isn't so unreasonable?
       
 (DIR) Post #AUbr2CmbfF1Bf2yXo0 by wizard@xyzzy.link
       2023-04-13T12:44:20.753499Z
       
       0 likes, 0 repeats
       
       @ro aha, seems sohttps://github.com/canonical/k8s-dqlitei bet this works way better than etcd tbh
       
 (DIR) Post #AUbudQsUWRD9x8AfS4 by ro@raru.re
       2023-04-13T13:07:31Z
       
       0 likes, 0 repeats
       
       @wizard oh interesting! i don't have much experience with etcd, what's the issue with it?
       
 (DIR) Post #AUbudRT0KgPzmNFpMe by wizard@xyzzy.link
       2023-04-13T13:24:38.241487Z
       
       0 likes, 0 repeats
       
       @ro honestly i hardly know much about the details, but in my experience etcd hits the disk really hard, since there's not really any in-memory caching it does. etcd isn't transactional either, and data corruption is pretty common with it in my experience. HA can resolve that for sure but even in smaller HA scenarios it can cause weirdness.
       
 (DIR) Post #AUbumVeo7dn3Nj6Bmq by wizard@xyzzy.link
       2023-04-13T13:26:19.388318Z
       
       0 likes, 0 repeats
       
       @ro i only ever ran a small scale 8 node kubernetes cluster on a bunch of intel NUCs, and i'm sure etcd would run way better on actual enterprise grade hardware. But for just slapping a kubernetes cluster wherever, sqlite seems like a way more logical choice.
       
 (DIR) Post #AUbvJCrJ2FCCdTXIBc by ro@raru.re
       2023-04-13T13:27:41Z
       
       0 likes, 0 repeats
       
       @wizard oh, for some reason i thought etcd is more config stuff and for transactions you'd use redis or something?
       
 (DIR) Post #AUbvJDUIhGO6aPmQy0 by wizard@xyzzy.link
       2023-04-13T13:32:09.531303Z
       
       0 likes, 0 repeats
       
       @ro it is for config stuff, but a kubernetes cluster modifies its config constantly and unendingly. any state change that happens with a kubernetes cluster, which can be as simple as a pod changing hosts, is reflected in the cluster's data store.i think you can use redis as a data store too, kubernetes is flexible with that stuff.
       
 (DIR) Post #AUbvShRShHUq5syTB2 by wizard@xyzzy.link
       2023-04-13T13:33:57.605939Z
       
       0 likes, 0 repeats
       
       @ro iirc there was about 200kbps-1mbps of disk activity on my etcd nodes when that cluster i ran was idling. it adds up fast!