What happens when my Rius workspace goes over its limits?

Rius refuses new members, workspaces and alert rules past your plan’s limit, and refuses a request above the workspace ingest rate with HTTP 429. Extra spans beyond your monthly allowance are not refused, they are billed as overage. After a downgrade that leaves you with too many workspaces, you choose which ones stay, and the rest are paused with their data still readable. Plans, limits and billing are covered in Plans and billing.

How it works

Rius has two kinds of limits, and they behave differently.

Hard limits on members, workspaces and alert rules. Inviting a member, creating a workspace or creating an alert rule past the limit is refused. The console shows “limit reached” with a link to Plan & usage.

Ingest rate per workspace. Each workspace has its own spans-per-minute limit. Above it, Rius refuses the whole request with HTTP 429 and a Retry-After header of up to 60 seconds. None of that request’s spans are stored unless the exporter sends the request again. A higher plan raises the rate.

Monthly span allowance. This is not a hard limit. Spans beyond your plan’s allowance are billed at the plan’s overage rate, prorated per span, and appear on the next monthly invoice. For the cost side of this, see Track LLM and agent cost.

Over the workspace limit after a downgrade. A downgrade is scheduled for the end of the current period, and until then your current plan and its limits apply. If the new plan allows fewer workspaces than you have, an admin has 48 hours to choose which ones stay. Until someone chooses, every workspace keeps ingesting.

Set it up

1. Open Settings → Plan & usage. It shows a meter for each limit, the month’s spans against your allowance and, during the trial, the date the trial ends. Only organization admins manage billing.
2. Check your plan against the table above before you invite members, add workspaces or create alert rules. If a request is refused with “limit reached”, use the link to Plan & usage and change plan.
3. If your exporter gets HTTP 429, wait for the Retry-After period and send the request again, or move to a plan with a higher ingest rate. A request that is refused is not stored. To send fewer traces, RIUS_SAMPLE_RATE sets head sampling of root traces, so a value below 1.0 sends fewer of them.

export RIUS_SAMPLE_RATE=0.5

4. After a downgrade that leaves you over the workspace limit, open Plan & usage and find Workspaces · which ones stay. Tick exactly as many workspaces as the new plan allows. Each row shows the workspace’s region and creation date. Click Keep the selected workspaces.
5. You can change your choice until the 48 hours are up. Once you choose, the workspaces you kept go on ingesting and the rest are paused right away. Their ingest is refused with HTTP 402, and their data stays readable.
6. To resume paused workspaces, upgrade, or delete workspaces until the organization is back within its limit. Either one resumes every paused workspace.

To keep sending traces from your application, the SDK setup stays the same whichever plan you are on:

import rius
 
client = rius.init(
    api_key="ri_...",
    service_name="my-agent",
)

What you get

• Clear refusals. New members, workspaces and alert rules past the limit are refused with “limit reached” and a link to Plan & usage, so nothing is silently dropped.
• Overage instead of a cutoff. Spans past your monthly allowance keep flowing and are billed per span at your plan’s overage rate. If you change plan mid-month, spans before the change are billed at the old plan’s rate and spans after it at the new plan’s rate.
• A bounded ingest rate. Each workspace has its own limit, so a busy workspace does not use up the rate of another.
• A safe downgrade. Downgrades take effect at the end of the period, every workspace keeps ingesting until you choose, and paused workspaces stay readable. If nobody chooses within 48 hours, the oldest workspaces stay and the rest are paused.
• Trial behavior. An organization on the trial keeps every workspace it already has and is only refused new ones past the limit. When the trial ends, ingest is refused with HTTP 402 until you subscribe, the spans sent in the meantime are dropped, not buffered, and everything already stored stays readable.

When something else fits

• If you sit near the ingest rate or the workspace limit, Growth and Pro raise both. If you are already on Pro, write to help@glassflow.ai about an Enterprise plan.
• For larger span volumes than the listed plans cover, an Enterprise plan is arranged through help@glassflow.ai.
• If you only need to see usage, the meters on Plan & usage show each limit without any change to your plan.

Related