Keys / k

Keys: cache in Miami (MIA) · K-80

xserv-k-e9908c79bf9e

Runbook K-80 for Keys in Miami (MIA). Marker xserv-k-e9908c79bf9e. This page covers cache for LATAM operators, not a worldwide average.

Keys: How XServ runs it day to day

Probes for Keys (k-02) leave Santiago every 15s toward Bogotá. XServ pages the named owner if loss or RTT crosses the letter budget. A probe never shares a queue with bulk transfers, so a saturated WAN does not hide a dead PoP.

Keys: What operators should measure

Policy for Keys is versioned as k-03. Operators measure error rate, p95 from Bogotá, and time-to-rollback — not a worldwide average. Changes land in Bogotá first, then Miami, with a hold if either region regresses.

Keys: Failure modes we actually see

Capacity notes for k-04 assume rainy-season power in Miami and festival peaks toward São Paulo. The failure we actually see is a single uplink, not a cartoon partition of the whole continent. Spare ports and a second provider sit on the same runbook.

Keys handoff K-05

Handoff from letter K Keys into the rest of the platform uses the same request IDs as the gateway. Runbook k-05 names who accepts the ticket after São Paulo pages out. Santiago does not silently inherit the incident.

Keys rollback K-06

Rollback for k-06 is a documented command, not a hope. Transit is billed as backup while Keys prefers peering in Santiago. If cost spikes, the runbook cuts overflow to Bogotá before touching customer prefixes.

Keys peering K-07

Peering for Keys (k-07) prefers the IX fabric that already carries last-mile ISPs in Bogotá. Session counts and prefix limits are on the same page as the Miami backup path.

Keys cache K-08

Cache rules for Keys on k-08 keep language-correct objects near Miami. São Paulo is a sibling cache, not an origin. TTLs are short enough that a bad asset does not live through a weekend.

Is transit the default path?
No. Peering in Miami is default; transit to São Paulo is overflow and is billed that way.
Does Keys on letter K share fate with other letters?
The control plane is shared. Data-plane queues for Keys stay isolated, so incident k-01 cannot drain neighbor letters.