Does Pinecone have an SLA? The 99.95% uptime standard, its 10/25/50% credits, and the 60-minute ticket rule that decides every claim

When a vector database goes down, the failure rarely looks like a database failure. It looks like a chatbot that stops answering, a search box that returns nothing, or a RAG pipeline that quietly falls back to worse results. The question that reaches the platform or FinOps lead afterwards is the usual one — *does Pinecone owe us anything?* — and Pinecone's answer turns less on how long the outage ran than on what your team did in its first hour.

Does Pinecone have an SLA with service credits?

Yes — but only for customers on a qualifying Enterprise Advance Commitment. Pinecone's Database Service Level Addendum (last updated 1 October 2026) commits to commercially reasonable efforts to keep its Database Services available at least 99.95% of the time in each calendar month. If a month falls short, the credit is 10%, 25% or 50% of that month's database fees, applied to future amounts due.

Two conditions decide almost every claim. You must have opened a support ticket within 60 minutes of the start of the downtime, and you must then file a separate "Request for SLA Credit" ticket by the end of the following calendar month. Pay-as-you-go and standard plan customers are outside the Addendum entirely, so no credit is owed on those plans however long an outage runs.

What does the Pinecone SLA actually promise?

Source: Pinecone's published Database Service Level Addendum, read on 9 October 2026; Ontracko's Pinecone SLA profile carries the same tariff.

How much downtime does Pinecone's 99.95% allow?

In a 30-day month (43,200 minutes), 99.95% leaves about 21.6 minutes of downtime; in a 31-day month, about 22.3 minutes. Beyond that, Pinecone's table prices the shortfall like this:

Unlike some tariffs we have read, Pinecone's three bands meet cleanly: every missed month lands in a priced band. For the general arithmetic, see what a 99.9% SLA actually allows and how SLA uptime is measured.

Why does the 60-minute ticket matter so much?

Because it comes first, and nothing later can repair it. To be eligible for a credit, the Addendum requires that you submitted a Support Ticket within 60 minutes of the start of the Downtime, reasonably assisted Pinecone's investigation, and followed its directions in response. An outage that nobody ticketed in its first hour cannot be claimed, however well the paperwork is done a month later.

That is a very different shape from most cloud SLAs, where the only clock is a 30- or 60-day filing window. In practice it means the claim is decided by your incident runbook, not by your finance team: whoever is on call when queries start failing has to open a Pinecone ticket inside the hour, even while they are still diagnosing. Our SLA claim deadlines by vendor roundup has the full comparison.

What doesn't count as Pinecone downtime?

The Addendum excludes time caused by any of its Exceptions: your breach of the Order or the Agreement; failure to configure and use the services as documented; failures of your own systems or connections; failures of, or issues with, Cloud Providers or Model Providers; force majeure; Pinecone's authorised suspension of access; and maintenance Pinecone announces in advance in the console or on its status page.

The cloud-provider exception deserves attention, because Pinecone indexes run on AWS, Google Cloud and Azure regions. If a hyperscaler incident takes your index down, the minutes are, on the document's wording, excused — check whether the underlying cloud's own SLA pays instead (see our guides for AWS, GCP and Azure).

Two more filters remove claims quietly:

• The replica rule. Pod-based and Dedicated Read Nodes indexes must have kept at least two replicas throughout the calendar month. A single-replica index for even part of the month disqualifies that month.
• The status page is not proof. The Addendum says the "uptime" shown on Pinecone's status page is informational and does not equal the Monthly Uptime Percentage. Use the Pinecone status page to confirm an incident happened; build the claim on your own request logs.

How to claim a Pinecone SLA credit

1. Confirm you are covered: you need Database Services under a qualifying Enterprise Advance Commitment (an Enterprise plan, a 12-month-plus Order, and an upfront payment against future consumption).
2. During the incident, open a Pinecone Support Ticket within 60 minutes of the downtime starting, then assist the investigation and follow Pinecone's directions.
3. Confirm the index met the replica rule: at least two replicas all month for pod-based or Dedicated Read Nodes indexes.
4. Rule out Exceptions — above all, a cloud-provider or model-provider failure, your own misconfiguration, or announced maintenance.
5. Pull minute-by-minute request logs showing the failed connections, redacting any personal or confidential information, and compute the Monthly Uptime Percentage for the month.
6. Open a separate Support Ticket with "Request for SLA Credit" in the subject line by the end of the calendar month after the downtime, including the event description, logs, affected end users or customers, the dates, times and duration, and what you did to resolve it at the time.
7. Check the credit lands against future amounts due for Database Services. It has no cash value and expires with the Order.

A worked example

Say your Database Monthly Fees are $8,000, and in a 30-day month one production index failed connections for a cumulative 90 minutes of qualifying, non-excused downtime — and your on-call engineer opened a ticket 20 minutes in. Uptime is (43,200 − 90) ÷ 43,200 ≈ 99.79%, inside "less than 99.95% to 99.0%", so the tier is 10% and the credit is 10% × $8,000 = $800.

Now say the same outage ran only 20 minutes. Uptime is (43,200 − 20) ÷ 43,200 ≈ 99.954% — still at or above the 99.95% standard, so nothing is owed. And if the 90-minute outage had happened with nobody opening a ticket in the first hour, the answer would also be nothing. The credit calculator on the Pinecone SLA page runs the same table against your own spend; our guide to calculating any SLA credit walks through the method.

Frequently asked questions

Does Pinecone serverless have an SLA?

Serverless indexes are covered Database Services under the Addendum, but only for customers on a qualifying Enterprise Advance Commitment. Ordinary pay-as-you-go serverless usage has no published SLA and no credit.

What is Pinecone's uptime guarantee?

A 99.95% monthly availability standard for Database Services, excluding Exceptions, on commercially reasonable efforts — backed by service credits of 10%, 25% or 50% of the month's database fees.

Am I owed a credit if Pinecone was down for two hours?

On a qualifying Enterprise Advance Commitment, two hours of qualifying downtime in a 30-day month is about 99.72% uptime, which pays 10% — provided you opened a support ticket within 60 minutes of the start, the downtime was not an Exception such as a cloud-provider failure, and you file the separate credit request on time. Without a qualifying commitment, no credit is owed.

How long do I have to claim a Pinecone credit?

Two clocks apply. The incident ticket must be opened within 60 minutes of the downtime starting; the "Request for SLA Credit" ticket must be filed by the end of the calendar month after the month of the downtime.

Can I use Pinecone's status page as evidence?

Only as corroboration. The Addendum states that the uptime shown on the status page is informational and does not equal the Monthly Uptime Percentage, and it asks for your own request logs documenting the errors.

Does Pinecone refund money for outages?

No. Service Level Credits are applied to future amounts due for Database Services, have no cash value, are non-transferable, expire with the applicable Order, and are the sole and exclusive remedy.

Does an AWS or GCP outage that takes down Pinecone count?

Not under Pinecone's SLA. Failures of, or issues with, Cloud Providers are an express Exception, so those minutes are excluded from Downtime.

Methodology & caveats

Every term above — the Enterprise Advance Commitment scope, the 99.95% Availability Standard and its "commercially reasonable efforts" wording, the per-index, full-minute Downtime definition, the uptime formula, the Exceptions list (including Cloud Providers and Model Providers), the 10/25/50% table, the two-replica rule, the 60-minute ticket requirement, the "Request for SLA Credit" filing and its deadline, the status-page disclaimer, and the credit-form and sole-remedy terms — is transcribed from Pinecone's published Database Service Level Addendum (last updated 1 October 2026), read on 9 October 2026. The tariff, standard and deadlines match the August 2024 version our vendor profile was first built from; the October 2026 revision adds Model Providers to the Exceptions and extends the replica rule to Dedicated Read Nodes. Downtime minutes are our arithmetic on 30- and 31-day months. Pinecone may revise the Addendum, and your Order may vary these terms; where it does, it controls. Ontracko scores status-page incident duration, which is evidence that an event occurred, not the Monthly Uptime Percentage Pinecone computes — which is why the claim needs your own logs. For context across vendors, see which SaaS vendors actually pay SLA credits, the OpenAI and Anthropic SLA guide, why SLA credit claims get denied, the reliability rankings and the SLA glossary.

*With Pinecone, a claim is won or lost in the first 60 minutes. Ontracko watches the Pinecone status feed for free, alerts you the moment an incident opens so the ticket goes in inside the hour, times the incident, and reminds you before the month-end filing deadline. Monitor Pinecone free — 8% only on recovered credits. See Pinecone live status or open the Pinecone SLA page.*