titan/atlas-iac
3
0
Fork 0
Code Issues Pull Requests 9 Actions Packages Projects Releases Wiki Activity

infra(core): join titan-24 as an amd64 worker for the image build leg

Browse Source 
The native amd64 hermes-agent image leg builds on titan-24. Worker membership
in this cluster is reconciled by the node-prefer-noschedule CronJob (kubectl
label), not Ansible, so add titan-24 there:

- clear_worker titan-24 amd64  -> node-role.kubernetes.io/worker=true + hardware=amd64
- a soft PreferNoSchedule guard taint (atlas.bstein.dev/sui-validator=true)
  mirroring titan-22's media guard, so routine pods do not crowd the
  out-of-cluster Sui validator that co-hosts titan-24. GPU workloads pinned to
  titan-24 by hostname are unaffected (PreferNoSchedule never blocks a pinned
  pod), and the amd64 build pod tolerates this taint explicitly.

Operator note: this reconciler does not manage cordons (owned by Ananke
recovery). titan-24 is on the recovery uncordon denylist, so the operator must
ensure titan-24 is uncordoned/schedulable before the first amd64 build.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BvMSXH8VH2tMWXanb8SJdf
This commit is contained in:
jenkins 2026-08-25 11:05:42 -03:00
parent ff2c003b4b
commit 8ddff9b626
1 changed files with 16 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
infrastructure/core/node-prefer-noschedule-cronjob.yaml
View File

@ -59,6 +59,7 @@ spec:
clear_worker titan-14 rpi4
clear_worker titan-18 rpi4
clear_worker titan-22 amd64
clear_worker titan-24 amd64
if k get node titan-22 >/dev/null 2>&1; then
k label node titan-22 \
@ -69,6 +70,21 @@ spec:
k taint node titan-22 atlas.bstein.dev/media-primary=true:PreferNoSchedule --overwrite=true || true
fi
if k get node titan-24 >/dev/null 2>&1; then
# titan-24 (amd64, RTX 3080) co-hosts the out-of-cluster Sui
# validator (docker-compose at /opt/sui-validator). It joins as a
# worker so the native amd64 hermes-agent image leg can build here,
# but a soft guard taint keeps routine pods from crowding the
# validator. GPU workloads pinned to titan-24 by hostname are
# unaffected (PreferNoSchedule never blocks a pinned pod); the
# amd64 build pod tolerates this taint explicitly.
k label node titan-24 \
atlas.bstein.dev/general-compute=last-resort \
atlas.bstein.dev/sui-validator=true \
--overwrite=true || true
k taint node titan-24 atlas.bstein.dev/sui-validator=true:PreferNoSchedule --overwrite=true || true
fi
if k get node titan-23 >/dev/null 2>&1; then
k label node titan-23 \
veles.bstein.dev/simulation=true \