$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
$ argoctl preflight --site dc1.yaml
  reading site file dc1.yaml
  bmc node-a ok   bmc node-b ok
✓ 2 nodes, 16 GPUs, fabric up
$ argoctl provision
  redfish: power cycle node-a node-b
  ironic: writing pinned ubuntu image
✓ OS image via Redfish  2/2
$ argoctl cluster up
  kubeadm: control plane on mgmt
  gpu-operator: installing (pinned)
✓ control plane ready
$ argoctl fabric verify
  ib0 up   ib1 up
  ib_write_bw node-a -> node-b
✓ RDMA path verified
$ argoctl qualify
  dcgm diag: level 3
  nccl all_reduce_perf  2 nodes x 8 GPUs
✓ all-reduce passed  16/16
$ argoctl status
  alerts routed to on-call
  acceptance report written
● operating  16/16 healthy
Mission

We believe a significant portion of enterprise AI workloads will shift from public cloud to hybrid / on-prem setups over the coming years, due to concerns around data sovereignty, cybersecurity and cost.
We are building the stack to let enterprises smoothly operate their own GPUs, and remove all the friction in migrating sensitive workloads away from public cloud to sovereign infrastructure.
Join us