Files
dingoandClaude Opus 5.5 680e7e5725 Start M8.1 inventory: stated facts, workload profile and collectors
Record the maintainer-stated host facts (R730xd ~128 GB, 32 GB workers,
automatic Harbor publish and Argo CD deploy, existing aevolution-db
Secret and its conflicts with the pinned version and role split), the
measured evo-server profile (peak RSS ~9.7 MiB, ~0.3% of a core) and a
local storage baseline. Add a read-only cluster collector that never
fetches Secret data and a by-hand storage latency probe outside the
Argo path. Cluster facts remain uninspected pending read-only access.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 11:57:08 -04:00

111 lines
3.9 KiB
Plaintext

# M8.1 storage latency probe. NOT part of any Argo path: apply by hand, once per
# candidate StorageClass, after replacing <STORAGE_CLASS> (left unresolved on
# purpose so this file cannot be applied unchanged). It measures the durability
# pattern the region host depends on, on a fresh PVC of that class:
# - 4 KiB synchronous writes (O_DSYNC), the PostgreSQL WAL-like cost;
# - write + fsync + rename + directory fsync of a 40 KiB object (the object
# store's publication pattern, ADR-0061);
# - 256 MiB sequential write with fdatasync (throughput and capacity sanity).
# Results are printed to the Job log. Afterwards delete the Job and the PVC
# (kubectl delete job,pvc,networkpolicy -l app.kubernetes.io/name=evo-disk-probe
# -n evolution-dev) and record the numbers in docs/infrastructure/INVENTORY.md.
# The image is the already-pinned evo-sim runtime (Ubuntu 24.04 coreutils); no
# new image is introduced.
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: evo-disk-probe
namespace: evolution-dev
labels:
app.kubernetes.io/name: evo-disk-probe
spec:
accessModes: [ReadWriteOnce]
storageClassName: <STORAGE_CLASS>
resources:
requests:
storage: 2Gi
---
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
name: evo-disk-probe
namespace: evolution-dev
labels:
app.kubernetes.io/name: evo-disk-probe
spec:
podSelector:
matchLabels:
app.kubernetes.io/name: evo-disk-probe
policyTypes: [Ingress, Egress] # No traffic in or out.
---
apiVersion: batch/v1
kind: Job
metadata:
name: evo-disk-probe
namespace: evolution-dev
labels:
app.kubernetes.io/name: evo-disk-probe
spec:
backoffLimit: 0
activeDeadlineSeconds: 600
template:
metadata:
labels:
app.kubernetes.io/name: evo-disk-probe
spec:
restartPolicy: Never
automountServiceAccountToken: false
securityContext:
runAsNonRoot: true
runAsUser: 10001
runAsGroup: 10001
fsGroup: 10001
seccompProfile:
type: RuntimeDefault
containers:
- name: probe
# Same repository and digest as deploy/kubernetes/dev (verified run 1399).
image: registry.aeternalisgames.com/aeternalis/aevolution@sha256:0c6c6e1ac28de2ce0648fd0992ca694963e303ef5c9414fe5ac668937136800d
command: ["/bin/bash", "-c"]
args:
- |
set -eu
cd /data
ns() { date +%s%N; }
echo "filesystem: $(df -hT /data | tail -1)"
t0=$(ns)
dd if=/dev/zero of=dsync.bin bs=4k count=1000 oflag=dsync status=none
t1=$(ns)
echo "dsync_4k_writes=1000 mean_us=$(( (t1 - t0) / 1000 / 1000 ))"
head -c 40960 /dev/urandom > payload.src
mkdir -p objects
worst=0; t0=$(ns)
for i in $(seq 1 200); do
s=$(ns)
dd if=payload.src of=objects/.tmp status=none conv=fsync
mv objects/.tmp objects/o$i
sync objects
e=$(( ($(ns) - s) / 1000 )); [ "$e" -gt "$worst" ] && worst=$e
done
t1=$(ns)
echo "object_publish_40k=200 mean_us=$(( (t1 - t0) / 200 / 1000 )) max_us=$worst"
t0=$(ns)
dd if=/dev/zero of=seq.bin bs=1M count=256 conv=fdatasync status=none
t1=$(ns)
echo "sequential_256m_ms=$(( (t1 - t0) / 1000000 ))"
rm -rf dsync.bin payload.src objects seq.bin
resources:
requests: {cpu: 100m, memory: 64Mi}
limits: {cpu: 500m, memory: 128Mi}
securityContext:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: true
capabilities:
drop: [ALL]
volumeMounts:
- {name: data, mountPath: /data}
volumes:
- name: data
persistentVolumeClaim:
claimName: evo-disk-probe