Content Routing resiliency

[BigLep]

Content Routing is a risk.
We don't want to have the situation where data isn't retrievable, and we point the finger at an underlying situation. That takes away from the narrative we're going for with FOC that "I store data with Filecoin and it works". Exposing/blaming the internal decisions we've made of using IPNI is similar to forcing a client to be educated on SPs.
I'm assuming we aren't building a new content routing solution between now and GA.
This s a placeholder with an estimate for any work we're going to do do monitor or communicate about the situation.

[BigLep]

2026-02-06 thoughts on the minimum we need here:

• Generally about any content routing system
  • We want to know if there's a problem with content routing before customers know
  • If there's a problem, we want to know who to escalate to, how to escalate to them, and to have assurance that we'll get help
• Ultimately we want to make sure content routing works for paid deals with approved SPs. Things that could get in the way:
  • Content routing system is down (should be caught by the above)
  • Individual SP is failing to make their content discoverable. This could be intentional or accidental.

How to determine if "IPNI is down"?

Specific things we can do:

1. Understand how IPNI monitors itself
   • There is https://probelab.io/ipfs/ipni/cid.contact/
   • What automated alerting do they have?
2. Look at signal from dealbot if we're seeing a lot more IPNI-related failures

• Note that dealbot right now is just querying filecoinpin.contact. It's not doing anything to check cid.contact. This could be added. Basically, first check filecoinpin.contact. If that looks good, then check cid.contact.

Alerting/escalation for IPNI system being down

Things we can do:

1. Meet with IPNI team and shipyard to understand how we page them if we need to.
2. Setup alarm if we see widespread consistent failure from dealbot with IPNI so we get a human in the loop.

Individual SPs aren't making their content discoverable

Things we can do:

• SP is intentionally not making their content discoverable
  • Dealbot will catch this if they are doing blanked IPNI disablement. Dealbot won't catch if the SP is being selective.
  • We could build a system that checks random pieces that have ipni indexing enabled to make sure their records show up in IPNI. I assume this is an improvement we would consider post GA.
• SP is unintentionally not making their content discoverable
  • We had a good example of this in https://filecoinproject.slack.com/archives/C07CGTXHHT4/p1770301489215909 is an example.
  • We could be querying https://filecoinpin.contact/providers/ to proactively look for errors that SPs may be having and to expose this on dealbot and alert
  • Dealbot will expose if there is issues with the pieces it has added. We can get slack notifications if we're seeing consistent failure for an approved SP.
  • We need to write up a runbook of what an SP can do to diagnose and fix the situation (i.e., make a runbook from https://filecoinproject.slack.com/archives/C07CGTXHHT4/p1770301489215909).

I know we need to commit what specifics we're going to do, but I think tackling most of the above is more like 3 days instead of 5. Does reducing the estimate seem good to you @rjan90 ?