One service degraded
Market data is running behind on one venue feed. Trading on that venue has stood down rather than acting on stale candles.
This page is not yet wired to live monitoring. The service states and history below are illustrative. Until the health endpoint is connected, treat an incident here as an example of how one is reported rather than as a current fact.
Decision engine
Eight-stage pipeline, all regions
Risk service
Mandatory veto path
Execution
Order routing and fills
Market data
Feed normalisation and backfill
HTTP API
api.trade-logx.com
Dashboard
Web application
Incident history
Every incident gets a write-up naming the cause and the change made, including the ones where the system behaved correctly. A halt that worked is still an event worth explaining, because from the outside it looks identical to a failure.
- · 41 minutes
Market data lag on one venue feed
A venue websocket delivered candles with increasing delay without disconnecting, so failover did not trigger. Gap detection caught the staleness and strategies on that venue stood down rather than trading old data. Failover now considers staleness, not only disconnection.
- · 18 minutes
Elevated API latency, EU region
A connection-pool exhaustion in the read path pushed p95 API latency above two seconds. Trading was unaffected — the decision path does not share that pool. Pool sizing is now derived from instance capacity rather than a fixed constant.
- · 6 minutes
Deliberate trading halt during a dependency failure
The risk service became unreachable from the execution workers. The system failed closed and stopped placing orders, which is the intended behaviour. Open positions retained their venue-resident protective orders throughout. Root cause was a mesh certificate rotation that raced its own reload.