feat(diariz): pilot stack on fv-ml1 GPU 3 — override (GPU pin, api port), vault-rendered .env, nightly idle-gated worker restart, fleet caddy site

2026-10-19 10:50:56 -07:00
parent 2929bfc60c
commit 9aecb677a8
10 changed files with 245 additions and 0 deletions
@@ -0,0 +1,13 @@
+# fv-ml1: the Diariz workers' nightly restart (~0300, idle-gated on GPU 3). Prime 2026-10-19.
+steps:
+  - {name: script, upload: {src: stacks/diariz/diariz-nightly-restart, dest: /usr/local/sbin/diariz-nightly-restart, mode: "0755"}, sudo: true}
+  - {name: service, upload: {src: stacks/diariz/diariz-nightly-restart.service, dest: /etc/systemd/system/diariz-nightly-restart.service, mode: "0644"}, sudo: true}
+  - {name: timer, upload: {src: stacks/diariz/diariz-nightly-restart.timer, dest: /etc/systemd/system/diariz-nightly-restart.timer, mode: "0644"}, sudo: true}
+  - name: Enable the timer
+    sudo: true
+    shell: systemctl daemon-reload && systemctl enable --now diariz-nightly-restart.timer && systemctl list-timers diariz-nightly-restart.timer --no-pager | head -3
+verify:
+  - name: Timer scheduled for 0300
+    sudo: true
+    shell: systemctl is-enabled diariz-nightly-restart.timer && systemctl show diariz-nightly-restart.timer -p NextElapseUSecRealtime --value
+    changed_when: "false"
@@ -0,0 +1,26 @@
+# fv-ml1: provision the Diariz pilot (kenhayward/Diariz @ 5a96cade, AGPL-3.0) beside Scriberr. Brokkr's request for
+# Prime (2026-10-19, thread 01M5AH9QCVW7ZQ6VVXGWQGFPRS); Prime 1035: GPU 3, with a nightly worker restart ~0300.
+# Source: a clean `git archive` of the pin, extracted to /opt/docker/src/diariz-5a96cad (owned by infra-ops).
+# The fleet override (stacks/diariz/docker-compose.override.yml) goes beside upstream's compose; /opt/docker/compose/diariz
+# is a symlink to that deploy dir. The .env is rendered from the vault separately (stacks/diariz/render-env.sh) and
+# piped to the host; it is never on disk on nh3-dev. This playbook does NOT build or start anything.
+vars:
+  src: /opt/docker/src/diariz-5a96cad
+steps:
+  - {name: stage source archive, upload: {src: /tmp/claude-1000/-home-lkraven-development-eshpfi-management/e0be18fd-0734-4786-b933-904f0a74d2a2/scratchpad/diariz-5a96cad.tar.gz, dest: /tmp/diariz-5a96cad.tar.gz, mode: "0644"}}
+  - name: Extract the pinned source (once)
+    sudo: true
+    shell: |
+      set -e
+      install -d -o infra-ops -g infra-ops -m 0755 {{src}}
+      runuser -u infra-ops -- tar -xzf /tmp/diariz-5a96cad.tar.gz -C {{src}}
+      test -f {{src}}/deploy/docker-compose.yml && rm -f /tmp/diariz-5a96cad.tar.gz
+    creates: /opt/docker/src/diariz-5a96cad/deploy/docker-compose.yml
+  - {name: fleet override, upload: {src: stacks/diariz/docker-compose.override.yml, dest: /opt/docker/src/diariz-5a96cad/deploy/docker-compose.override.yml, mode: "0644"}}
+  - name: /opt/docker/compose/diariz -> the deploy dir
+    sudo: true
+    shell: ln -sfn {{src}}/deploy /opt/docker/compose/diariz && ls -ld /opt/docker/compose/diariz
+verify:
+  - name: Source pinned and the override in place
+    shell: cd {{src}}/deploy && ls docker-compose.yml docker-compose.override.yml && grep -c device_ids docker-compose.override.yml
+    changed_when: "false"