Keys, quotas and limits
How Z2 access keys, shared capacity, per-bucket quotas and object limits behave — and what the S3 API does not support.
Z2 is deliberately simple: one plan, one key, one capacity pool, buckets carved out of it. Knowing the boundaries avoids surprises.
Access keys
- One key per storage plan. It has owner rights on every bucket in that plan.
- Show credentials reveals the Key ID and Secret. Treat the secret like a password.
- Rotate key creates a new pair and revokes the old one immediately. Every client using the old key stops working at that moment; update your applications first, then rotate during a maintenance window.
Keys are per plan, not per bucket. If two teams need isolated credentials, give them separate storage plans.
Capacity and quotas
| Rule | Behaviour |
|---|---|
| Plan capacity | The plan's total bytes are shared by all its buckets |
| New bucket quota | Receives the remaining unallocated capacity as a hard quota |
| Minimum to create | At least 1 GiB must remain unallocated |
| Over quota | Writes are refused with 403 QuotaExceeded; no overage billing |
| Resizing a bucket | Not exposed in the console. Rebalance by creating and deleting buckets |
Because the first bucket takes all remaining capacity, a single-bucket plan is simple. Multi-bucket plans need a deliberate split: create the first bucket, then delete and recreate it if you need to leave room for others, or start with the smaller bucket.
The console's Capacity card shows Stored, Allocated to buckets and Unallocated, which is the clearest way to see where the space went.
Object limits
| Limit | Value |
|---|---|
| Maximum object size | 1 GiB |
| Console single upload | 1 GB |
| Object key length (console) | 900 characters |
| Bulk delete per request | 1,000 keys, or a prefix ending in / |
| Multipart session lifetime | 24 hours |
| Presigned download expiry (console) | 300 seconds |
For larger files, split them or use the S3 API with multipart upload. Multipart sessions that are never completed are cleaned up after 24 hours.
Bucket names
- 3–63 characters
- Lowercase letters, digits and hyphens
- Start and end with a letter or digit
- Cannot be renamed after creation
What is not supported
- Versioning — overwriting an object replaces it.
- ACLs and public bucket policies — access is through your key or a presigned URL.
- Object tagging and object lock.
- Conditional requests (
If-Match,If-None-Match). - Server-side copy —
CopyObjectis implemented as a download and re-upload, so it is not free. - Virtual-hosted-style addressing — use path style.
- Bucket create/delete through S3 — use the console.
Overwrites delete the old object and upload the new one. There is a brief window during an overwrite where the object does not exist. If readers must never see a missing file, write to a new key and switch the reference.
Billing and performance
- No egress or request fees. Transfer and API calls are not metered.
- Throughput is bounded by the storage upstream, not by an artificial per-tenant limit. Very large parallel workloads will see contention.
- Listing can be throttled under heavy parallel load. Keep large sweeps serial with backoff rather than opening dozens of concurrent list operations.
Security checklist
- Keep bucket contents private by default; hand out presigned URLs with short expiries.
- Do not commit the access key. Load it from the environment or a secret manager on the machine.
- Rotate the key if it may have leaked, then audit what the old key could reach.
- Use one storage plan per trust boundary rather than sharing a key across teams.
- Remember that a presigned URL is a bearer token: anyone with the link can use it until it expires.
Next
- access key
- rotate
- quota
- limits
- object size
- versioning
Ready to use this
S3-compatible buckets your existing tools already speak.
- No egress fees
- Per-bucket quotas from the console

