Roger Oriol
58bba66f18
simplify all of platform engineer's crons into one single daily cron
2026-08-01 16:52:45 +02:00
Roger Oriol
33b1b2dc8f
fix deepseek pvc
2026-08-01 10:46:09 +02:00
Roger Oriol
27307ff683
fix deepseek pvc
2026-08-01 10:34:54 +02:00
Roger Oriol
475bf48fe4
fix deepseek deployment
2026-08-01 10:20:05 +02:00
Roger Oriol
b132aae09c
deploy deepseek v4
2026-08-01 00:24:26 +02:00
Roger Oriol
5fe0ee361a
add openwakeword deployment
2026-07-29 00:43:23 +02:00
Roger Oriol
fba742b34f
add kubernetes local host resolution to homeassistant
2026-07-29 00:29:20 +02:00
Roger Oriol
90a04b96a3
give whisper pod more memory
2026-07-29 00:15:29 +02:00
Roger Oriol
8c5c884b4d
add stt and tts services
2026-07-29 00:07:10 +02:00
Roger Oriol
9eb425f7f5
merge
2026-07-28 23:02:08 +02:00
Roger Oriol
6970313822
fix homeassistant not being able to see scripts saved
2026-07-28 23:01:22 +02:00
Hermes Platform Engineer
9a230d03c9
Merge fix(homeassistant): reduce memory requests + clean up restartedAt annotation
2026-07-28 19:22:16 +00:00
Hermes Platform Engineer
71f724420e
Merge fix(homeassistant): memory reduction with restart annotation cleanup
2026-07-28 19:22:16 +00:00
Hermes Platform Engineer
6967925c4f
fix(homeassistant): add restartedAt annotation to trigger rollout and clean up stale Pending pod
...
The old ReplicaSet 789f79897c left a Pending pod (home-assistant-789f79897c-plf7n)
because the nucbox node is at 95.9% memory. Adding restartedAt annotation to
trigger a fresh rollout so the new ReplicaSet takes over.
2026-07-28 13:11:08 +00:00
Hermes Platform Engineer
2f1ae311f7
fix(homeassistant): reduce memory requests to 256Mi, add node affinity for raspberrypi
...
Node roger-nucbox-evo-x2 is at 3.78% memory free causing home-assistant
pods to flap Pending/Running. Reduce resource requests to fit within
cluster capacity and prefer scheduling on raspberrypi (63% mem free).
2026-07-28 12:58:19 +00:00
7ca919b7c7
Merge pull request 'fix(homeassistant): pin image to 2026.7.4 to avoid cffi version mismatch' ( #28 ) from fix/homeassistant-pin-version into main
...
Reviewed-on: #28
2026-07-28 11:07:26 +02:00
Hermes Platform Engineer
d266086c1b
fix(homeassistant): pin image to 2026.7.4 to avoid cffi version mismatch with Python 3.14
2026-07-28 09:01:06 +00:00
Roger Oriol
d3798f47ac
fix homeassistant configuration.yaml not being writable
2026-07-28 00:31:23 +02:00
Roger Oriol
707ec39291
fix hermes api not being reachable
2026-07-26 17:19:20 +02:00
Roger Oriol
5bb1066d0d
fix hermes api not being reachable
2026-07-26 17:08:27 +02:00
Roger Oriol
575c7bf9e0
agents fix language and time zone
2026-07-26 14:59:58 +02:00
Roger Oriol
4222582a54
Merge branch 'main' of https://git.rogi.casa/roger/k3s-cluster
2026-07-26 14:38:02 +02:00
Roger Oriol
4f6255800b
enable api access for agents home manager and platform engineer
2026-07-26 14:37:37 +02:00
d1e86ce5ec
fix(prometheus): add restart annotation to trigger rollout and clean up stale pods
2026-07-26 13:26:46 +02:00
Platform Engineer Agent
d5e4e37358
fix(prometheus): bump memory limit from 2Gi to 4Gi to prevent OOMKilled
2026-07-26 10:44:19 +00:00
platform-engineer-agent
63557ef1b4
fix(argocd): delete argocd/apps/argocd.yaml (moved to argocd/argocd.yaml)
2026-07-26 07:04:51 +02:00
platform-engineer-agent
65bcb0c7a4
fix(argocd): remove argocd app from apps/ to avoid circular dependency
2026-07-26 07:02:15 +02:00
Roger Oriol
7af7ae66de
increase qwen3.6 models context available to 131k
2026-07-26 01:29:23 +02:00
Roger Oriol
5eaea8c387
fix home manager discord integration
2026-07-26 01:00:25 +02:00
Roger Oriol
8f24290f38
fix platform engineer daily cron
2026-07-26 00:27:38 +02:00
Roger Oriol
a7e78c18c3
fix pod restart cron
2026-07-25 20:19:27 +02:00
Roger Oriol
8caf49fdf6
fix home manager crons
2026-07-25 20:03:16 +02:00
Roger Oriol
842815d370
home manager crons fix
2026-07-25 00:54:46 +02:00
Roger Oriol
4694034a04
set qwen models parameters to allow for a higher context window
2026-07-24 21:30:01 +02:00
Roger Oriol
bc165d3cf1
change home-manager model to qwen3.6-35b-a3b
2026-07-24 20:37:43 +02:00
Roger Oriol
c982b33015
increase llamacpp models context window!
2026-07-24 20:11:30 +02:00
Roger Oriol
496e46076c
set platform engineer url to platform-engineer.rogi.casa
2026-07-23 23:57:41 +02:00
Roger Oriol
04bd4d8e5e
deploy both qwen3.6 versions
2026-07-23 22:10:44 +02:00
Roger Oriol
8ef9db1e37
fix llamacpp
2026-07-23 00:26:58 +02:00
Roger Oriol
70e4baf5e3
fix llamacpp image
2026-07-23 00:12:08 +02:00
Roger Oriol
0143ebefd9
new llamacpp service
2026-07-22 23:54:52 +02:00
Roger Oriol
5f7f1bd52a
set litellm master key
2026-07-22 22:03:40 +02:00
Roger Oriol
b58abda201
update gitea runner image
2026-07-22 21:32:09 +02:00
Roger Oriol
4307545436
drop master key from litellm general settings
2026-07-22 21:26:38 +02:00
Roger Oriol
84be8f2c36
merge
2026-07-22 21:20:27 +02:00
Roger Oriol
a8c7df67fe
increase litellm memory allowance
2026-07-22 21:17:51 +02:00
platform-engineer
097fe2e0e7
fix(prometheus): bump memory limit from 1Gi to 2Gi due to OOM
...
fix(litellm): add resource limits (512Mi/1Gi) to prevent unbounded memory consumption
2026-07-22 05:58:23 +00:00
Platform Engineer
6df0be81c9
fix(platform): fix high-restart pods - prometheus nodeSelector, litellm resources, gitea-runner limits
...
- Prometheus: change nodeSelector from hardware=high-memory (nonexistent label) to kubernetes.io/arch: amd64
Fixes OOMKilled cause - pod was being scheduled on wrong nodes.
- litellm-deployment: add resource limits (2Gi/500m requests, 4Gi/2000m limits) + pin image to v1.34.0
Fixes OOMKilled cause - no resource limits were set.
- gitea-runner: pin image to v0.12.0 from :latest + add resource limits (512Mi/250m requests, 1Gi/500m limits)
Addresses extreme restart count (281) caused by unbounded memory usage and rolling image updates.
2026-07-21 20:28:12 +00:00
a5291da0b2
fix(gitea): bump restartedAt annotation to clear stuck gitea-runner pods (281 restarts from DNS resolution failures)
2026-07-21 15:36:57 +02:00
platform-engineer
075bdd8ca3
fix(gitea-runner): pin image tag to v0.7.2, add resource limits (256Mi/100m req, 512Mi/500m lim) to prevent OOM restarts
2026-07-20 22:19:37 +00:00
