mattcary

Run of most of the tests, which ran a bunch then timed out:

Summarizing 48 Failures:
  [FAIL] E2E Test Suite [Driver: gcsfuse.csi.storage.gke.io] [Testpattern: Pre-provisioned PV (default fs)] volumes [It] [csi-skip-bucket-access-check] should store data using custom sidecar container image
  /usr/local/google/home/mattcary/local/gcs-fuse-csi-driver/test/e2e/specs/specs.go:310
  [PANICKED!] E2E Test Suite [Driver: gcsfuse.csi.storage.gke.io] [Testpattern: Dynamic PV (default fs)] volumes [It] [read-only] should fail when write
  /usr/local/google/home/mattcary/local/gcs-fuse-csi-driver/test/e2e/utils/iam_utils.go:351
  [PANICKED!] E2E Test Suite [Driver: gcsfuse.csi.storage.gke.io] [Testpattern: Dynamic PV (default fs)] istio [It] should store data with istio injected at index 0
  /usr/local/google/home/mattcary/local/gcs-fuse-csi-driver/test/e2e/utils/iam_utils.go:351
  [PANICKED!] E2E Test Suite [Driver: gcsfuse.csi.storage.gke.io] [Testpattern: Dynamic PV (default fs)] volumes [It] [fsgroup delegation] sh

mattcary

# This is run on a single host with 8 H200 GPUs. The GKE node pool creation command is:
# gcloud container node-pools create a3 \
#     --cluster a3 --location europe-west1 --num-nodes=1 \
#     --node-locations europe-west1-b \
#     --accelerator type=nvidia-h200-141gb,count=8 \
#     --machine-type=a3-ultragpu-8g

apiVersion: v1
kind: ConfigMap
metadata:

mattcary

# The system csi driver pod
$ kubectl get pod -n kube-system gcsfusecsi-node-sjdrk -o yaml
apiVersion: v1
kind: Pod
 <snip>
spec:
  containers:
  - args:
   <snip>
    image: us-central1-artifactregistry.gcr.io/gke-release/gke-release/gcs-fuse-csi-driver:v1.10.2-gke.0@sha256:dee682111037c222babd959717cc420e75985d938fade90f297dc254aec4c4f4

mattcary

## `03-pgbench-tbs.yaml` / SSD

### Setup

```
kubectl cnpg pgbench --dry-run \
  --db-name pgbench \
  --job-name pgbench-init-pgbench-tbs \
  --node-selector workload=pgbench \
  pgbench-tbs \

mattcary

## `01-pgdata-only.yaml`

Using n2-16 and standard-rwo (pd-balanced).

Regional cluster, so cross-zone n/w

### Setup

 kubectl cnpg pgbench --dry-run \
  --db-name pgbench \

mattcary

# Experiemnts

## `02-pgdata-wal.yaml`

Using n2-16 and standard-rwo (pd-balanced).

Regional cluster, so cross-zone n/w

### Setup

mattcary

apiVersion: apiextensions.k8s.io/v1
kind: CustomResourceDefinition
metadata:
  name: validations.mattcary.info
spec:
  group: mattcary.info
  names:
    kind: Validation
    listKind: ValidationList
    plural: validations

mattcary

We'll create a VM, attach a disk, mount it, write to it, then detach and watch
for fireworks.

```
gcloud compute instances create --machine-type n2-standard-4 poison-0
gcloud compute disks create --size 10G --type pd-standard p0-std
gcloud compute instances attach-disk --disk p0-std poison-0
gcloud compute ssh poison-0
$ ls -l /dev/disk/by-id
lrwxrwxrwx 1 root root  9 Oct  7 23:42 google-persistent-disk-0 -> ../../sda

mattcary

mattcary

# This produced an image like this:
# $ gcloud compute images describe snapshot-acc08d47-9a96-4c4b-a330-020e2c017fbc
# creationTimestamp: '2022-04-20T13:56:05.499-07:00'
# diskSizeGb: '12'
# id: '2238164906300788106'
# kind: compute#image
# labelFingerprint: 42WmSpB8rSM=
# name: snapshot-acc08d47-9a96-4c4b-a330-020e2c017fbc
# selfLink: https://www.googleapis.com/compute/v1/projects/mattcary-gke-dev2/global/images/snapshot-acc08d47-9a96-4c4b-a330-020e2c017fbc
# sourceDisk: https://www.googleapis.com/compute/v1/projects/mattcary-gke-dev2/zones/us-central1-c/disks/pvc-524a5b27-6130-422e-adc5-7ab5375fe81a