What it watches · 8 of 14
Pure Storage arrays
FlashArray and FlashBlade through Purity's REST API with a read-only token: controllers, blades, volumes, file systems, replication, Purity's own alerts, performance, and the PVCs each volume backs.
The clusters see their volumes as claims that are bound or not. They don't see the array behind them: a controller failing over, a failed drive, an ActiveCluster pod out of sync, a replication link down. Cairn OBS reads the arrays themselves, and links each array volume to the Kubernetes volume claim it backs.
Pure judges the array. Purity raises its own alerts, with its own thresholds, and Pure's support already knows them. Cairn OBS passes them on, and reads the few states that matter on their own: controllers, hardware, pods and links. There are no space or latency thresholds of its own.
Set it up
- On each array, make a user with the built-in Read-Only role. In Purity's GUI: Settings › Users, make the user, then Create API Token on it.
- On a host that can reach the arrays' management addresses on port 443, put each token in a file of its own, readable by the agent only, and list the arrays in its agent.yaml:
pure:
- url: https://fa-prod-01.example.org
token_file: /etc/obs-beacon/fa-prod-01.token
fingerprint: "AB:CD:…" # the array's self-signed certificate, SHA-256
- url: https://fb-prod-01.example.org
type: flashblade
token_file: /etc/obs-beacon/fb-prod-01.token
ca_file: /etc/obs-beacon/storage-ca.pem
| Key | What it's for |
|---|---|
type |
flasharray (the default) or flashblade |
url |
The management address |
token_file |
The API token, alone in the file |
ca_file, fingerprint |
Check the certificate against a CA, or pin it. Otherwise the system's CAs |
every |
Default 60s |
name |
What it's kept under. Default its host name |
Purity//FA 6.0 or Purity//FB 3.0 and later (REST 2). After logging in with the token, the agent only reads.
What's read
| FlashArray | FlashBlade | |
|---|---|---|
| The array | Model, version, capacity, space used, data reduction | The same |
| Its nodes | Controllers: offline or not ready is down; primary or secondary | Blades: down, evacuating, evacuated |
| Hardware | Hardware and drives not healthy, parity under 1.0 (rebuilding): degraded, saying which | The same |
| Storage | Volumes, with size, use and data reduction | File systems and buckets, with protocols, objects and quota |
| Replication | ActiveCluster pods (resyncing, offline, mediator out of reach) and array connections | File system and bucket replica links, and array connections |
| Alerts | Purity's open alerts, on the record they're about. A critical one makes it degraded | The same |
| Performance | Reads and writes a second, latency, bandwidth, space; hourly, kept 400 days | The same |
Each array is a source of its own on Clusters, "fa-prod-01 · read by mgmt1", with its numbers and a day, 30 days or a year of them.
Arrays and clusters
Pure's CSI driver and Portworx name each array volume after the Kubernetes PersistentVolume it backs. Cairn OBS links the two both ways:
- A volume claim's page shows its array volume and the array's state.
- An array volume's page shows which claim it is, in which cluster.
- An array in trouble says how many claims its volumes hold, and where: "Its volumes hold 312 PVCs, in prod, cluster-01".
A volume in an ActiveCluster pod is on both arrays, so its claim links to both.
Status
Tested against a fake built from Pure's published API and client field names, not yet against a real array. If you run it, tell us how it goes.