Deploy swarm-wkr-01/02 worker VMs #16

New issue

Closed

opened 2026-07-07 03:48:52 +00:00 by claude-bot · 0 comments

claude-bot commented

2026-07-07 03:48:52 +00:00

Collaborator

Clone 2 VMs from template 9000 (full clone, not linked):
- swarm-wkr-01, vmid 221, --target node03
- swarm-wkr-02, vmid 222, --target node05
- 8 vCPU / 8GiB RAM (matches current manager specs)
- ipconfig: 10.10.10.22{1,2} (vlan10 mgmt, gw 10.10.10.1),
  10.10.35.22{1,2} (vlan35 Ceph client), 10.10.30.22{1,2} (vlan30 ingress)
- nameserver 10.10.10.1, searchdomain coppercore.net
- confirm disk landed on Ceph (rbd -p vm-data ls), start VMs
- install qemu-guest-agent manually post-boot (known template gap)
Run ansible-playbook swarm-provision.yml --ask-vault-pass from
ccnet-swarm-nodes-runbook/ against the full inventory (managers + new
workers in the same run, required for the worker join-token hostvar
lookup to work).
Verify: docker node ls shows 5 Ready nodes (3 manager/1 leader, 2
worker); CephFS mounted on both new hosts; Netdata parent dashboard shows
2 new streaming children; re-run playbook to confirm idempotency.
Redeploy the 3 stacks (docker stack deploy -c <file> <stack>) so the
already-loosened placement constraints take effect, and optionally
docker service update --force <service> to actually rebalance existing
tasks onto the new workers.

Full command-by-command plan: /home/copper/.claude/plans/that-sounds-good-to-dreamy-minsky.md
(local session plan file, not in the repo).

Ansible/config side is done (main@b6c7947): swarm_workers inventory group, group_vars, updated swarm-provision.yml, and node.role==manager constraints already dropped from Forgejo/Nextcloud/SearXNG stacks. What's left is the actual infra: 1. Clone 2 VMs from template 9000 (full clone, not linked): - swarm-wkr-01, vmid 221, --target node03 - swarm-wkr-02, vmid 222, --target node05 - 8 vCPU / 8GiB RAM (matches current manager specs) - ipconfig: 10.10.10.22{1,2} (vlan10 mgmt, gw 10.10.10.1), 10.10.35.22{1,2} (vlan35 Ceph client), 10.10.30.22{1,2} (vlan30 ingress) - nameserver 10.10.10.1, searchdomain coppercore.net - confirm disk landed on Ceph (`rbd -p vm-data ls`), start VMs - install qemu-guest-agent manually post-boot (known template gap) 2. Run `ansible-playbook swarm-provision.yml --ask-vault-pass` from ccnet-swarm-nodes-runbook/ against the full inventory (managers + new workers in the same run, required for the worker join-token hostvar lookup to work). 3. Verify: `docker node ls` shows 5 Ready nodes (3 manager/1 leader, 2 worker); CephFS mounted on both new hosts; Netdata parent dashboard shows 2 new streaming children; re-run playbook to confirm idempotency. 4. Redeploy the 3 stacks (`docker stack deploy -c <file> <stack>`) so the already-loosened placement constraints take effect, and optionally `docker service update --force <service>` to actually rebalance existing tasks onto the new workers. Full command-by-command plan: /home/copper/.claude/plans/that-sounds-good-to-dreamy-minsky.md (local session plan file, not in the repo).

claude-bot

2026-07-07 03:48:52 +00:00