What data for this category looks like
The full coil-level or piece-level rolling process generates data for this category. Data sources include slab/billet batch information, furnace temperature control data, mill automatic control parameters, online thickness and width measurement equipment, surface defect detection systems, and MES systems. Daily documents cover rolling plans, heating schedules, rolling regulations, plate shape and thickness records, surface quality reports, and quality certificates. Fields include batch number, thickness value (mm), surface defect type, furnace temperature (℃), rolling speed, and others. Data updates in real time per coil or piece. Single batch data links multiple source collection items, forming a mixed dataset of structured and unstructured data.
What constraints these characteristics impose on deployment and upgrade
Multi-source real-time updates at coil level require deployments to prioritize high-concurrency writes and low-latency queries. Multi-source data association storage needs require database configurations to support nested documents and batch writes, to avoid broken associations in batch data. High-frequency real-time data creates storage and index pressure, so cache and storage expansion plans must be planned in advance. Errors in steel rolling directly relate to production line safety and customer complaints, so deployments must ensure high availability to prevent traceability interruptions from single-node failures. Deployments in factory intranet environments require configuring intranet-compatible access policies to avoid security risks from public network exposure.
How to configure
This page's configuration applies to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
MONGODB_URI | Replica set mode, connection pool size set to 32 | Needs to store multi-source inspection data associated with coil levels; replica sets ensure high availability, connection pool matches concurrent requirements of real-time writes |
REDIS_URL | Enable persistence, memory quota set to 8000 MB | Caches batch query results and real-time inspection data streams; persistence prevents session data loss on restart, quota matches cache volume of single-coil data |
STORAGE_S3_ENDPOINT | Configure domestic acceleration nodes | Stores surface quality reports, quality certificates and other documents; accelerates image pulling and document reading speeds, resolves issues of excessively long offline image pulling time |
DB_MAX_LINK | Set to 64 | Adapts to real-time data connection counts reported by multiple mill devices simultaneously, prevents collection interruptions caused by exhausted connections |
PG_URL | Structured data connection pool set to 16 | Stores structured documents such as rolling plans and heating schedules; connection pool adapts to concurrent query requirements |
FE_DOMAIN | Bind to factory intranet domain name | For production line terminal access, avoids security risks from public network exposure, matches intranet deployment environment of rolling workshops |
Three common mistakes
- Phenomenon: Image pull timeout when executing
docker compose up, logs show connection timeout errors. Cause: Domestic accelerated mirror sources are not configured, pulling official mirrors directly causes excessive network latency. - Phenomenon: Inconsistent batch query results after deploying multiple Kubernetes replicas. Cause: Shared
REDIS_URLcache is not configured, each replica caches independent batch data, leading to conflicting query results. - Phenomenon: Traceability queries for real-time inspection data return empty values. Cause:
MONGODB_URIis not configured for replica set mode, single-node failure prevents data from being read.
How to confirm configuration is complete
- Run container status check commands to confirm all service containers are in running state.
- Submit a query request for any rolling batch, verify that returned results include associated documents such as rolling plans and surface quality reports.
- Simulate multiple concurrent batch query requests to confirm no connection exhaustion-related errors occur.
- Execute the pull command after configuring image acceleration, confirm download time meets intranet deployment expectations.
The values provided are common starting points and should be measured against the reader's own samples.