Skip to main content

Release Notes

v1.4.10 · 2026-07-23​

Fixed

  • Host monitoring disappeared after upgrading (introduced in 1.4.8): since 1.4.8 the collection configuration defaults to the UI Data Sources page, but the upgrade migration missed the "host" type. Deployments that configured host monitoring via ENV_PROMETHEUS_HOST ended up with an empty host panel and no host scrape job in Prometheus at all — with no error anywhere. Upgrading to this version restores it automatically; no manual data source needs to be added.
  • Data sources created through the API defaulted to disabled: when the request body omitted enabled, the source was treated as disabled, so bulk-importing data sources via the API left everything uncollected until each one was toggled on manually. They are now enabled by default; pass false explicitly to disable.
  • Kafka / MongoDB / MySQL data sources created through the API had no purposes: when purposes was omitted the source ended up with none, and collection only picks up sources whose purposes include "metrics" — so the source was created, visible in the list and switched on, yet never collected. Default purposes are now applied.
  • The same MongoDB registered twice: the built-in slow-query diagnostic source and the automatically imported collection source pointed at the same address and appeared as two rows; deleting one of them brought it back after a restart. They are now merged into a single row with both purposes.
  • The purposes of built-in data sources were reset on every restart.
  • A failed "Test Connection" showed only a red icon with no text (for example, wrong Elasticsearch credentials should show the 401 detail).

New

  • Collection Status (Data Sources page): every data source shows "Collecting / No Data / Stopped / Not Collected" directly; a red status reveals the raw failure reason and offers copy-ready troubleshooting commands.
  • Post-deployment self-check (Overview page): summarises six areas — metric collection, host monitoring, HAP service logs, tracing, alert notifications and slow-query diagnostics — with actionable advice when something is wrong.
  • "Auto-imported" tag: data sources imported from ops.yaml environment variables are labelled, explaining where they came from and how to maintain them from now on.
  • Duplicate detection: when adding a data source that points at an instance you already registered (comparing address, database name, account and db index), the platform warns and offers to edit the existing source's purposes instead; same address with a different database or account is unaffected.
  • Alert rules with no notification channel are marked "No notification" in the list, explaining that firing only writes alert history.
  • The startup log now prints the key variables side by side, distinguishing the platform's own database from the monitored one, and the HAP-side Loki variable from this platform's identically named one.

Changed

  • The slow-query rule description now lists all four conditions — execution time, documents examined, how often the same query appears, and the collection interval — and shows a checklist when there is no data. Previously only the first two were shown, which made a working feature look broken.
  • The single source of truth for collection configuration is now the UI Data Sources page by default (ENV_PROMETHEUS_SD_MODE defaults to http); the middleware environment variables in ops.yaml are imported once, on the first startup.
  • Exporters are now supervised and restarted automatically, and a collection self-check table is printed about 40 seconds after startup.
  • Deployment example: data volumes are consolidated under volume/data/mdis/; every service gets restart: always; middleware connection variables are commented out by default (only needed for fresh installations); the Elasticsearch credential format is corrected to "username:password", and the Elasticsearch shipped with the HAP private deployment uses the username md.

Upgrade Notes

  • To upgrade, change the tag in x-ops-image at the top of ops.yaml, then run docker compose -f ops.yaml pull and up -d. Data volumes are untouched; monitoring history and alert configuration are preserved.
  • Upgrading from 1.4.7 or earlier: on the first startup the middleware and host environment variables in ops.yaml are imported into UI data sources automatically; from then on the UI is authoritative and editing ops.yaml has no effect.
  • Do not change the data volume paths to match the new documentation: the volume/data/mdis/ layout is for fresh installations only. Changing it on a running instance mounts empty directories and looks like data loss (the data is still at the original paths).

v1.4.6 · 2026-07-17​

  • Trace service names now strip any namespace suffix instead of only matching the default namespace.

v1.4.5 · 2026-07-16​

  • Data sources are enabled by default after creation, consistent across all types.
  • A failed ops-mongo connection now retries in the background with backoff instead of permanently disabling the alert subsystem — startup ordering no longer requires manual intervention.

v1.4.4 · 2026-07-16​

  • Prometheus supports size-based retention (retention.size).
  • The cluster deployment blueprint is aligned with the actual object-storage setup.

v1.4.3 · 2026-07-15​

  • Object storage compatibility fixes: Loki uses path-style, Tempo uses bucket_lookup_type for Tencent Cloud COS and Alibaba Cloud OSS.
  • Fixes related to appProtocol for istio tracing.

v1.4.1 · 2026-07-14​

  • Fixed MySQL / MongoDB monitors not decrypting passwords, which caused connection failures and false alerts.

v1.4.0 · 2026-07-13​

New

  • Built-in alerting subsystem, replacing the former Grafana alerting: seven monitoring types (base resources including custom PromQL, ports, SSL, MySQL, MongoDB, Kafka backlog, Kafka rebalance); a warning → alert → recovery noise-reduction state machine that only notifies on state changes; multiple channels including Feishu / DingTalk / WeCom / Email / custom Webhook; global alert history filterable by type / status / time.
  • Unified data source management: a new top-level Data Source page centrally manages hosts / middleware / K8s / Flink (multiple instances supported), with the left-hand menu displayed dynamically based on enabled data sources; "Test Connection" performs real authentication; credentials are AES-encrypted; optional http service discovery (adding a data source in the UI automatically brings up the exporter and registers it for scraping, with no need to change environment variables and restart).
  • MongoDB slow query support for sharded clusters: multiple instances are automatically aggregated for analysis.

Architecture changes

  • Single-image delivery: all components are merged into ops-allinone, distinguished by role via ROLE; you only need to pull / import one image.
  • Independent storage for alert data: a new ops-mongo container stores rules / state / history / channels / data sources, decoupled from the business database.

Upgrade notes

  • ops.yaml has undergone significant structural changes (single image + ROLE); regenerate it following the latest deployment docs.
  • A newly required setting, ENV_ALERT_CRYPTO_KEY (the credential encryption key), must be configured.
  • The former Grafana alerting / notification / silence screens have been removed; rebuild your rules in the new Alert Management section.

v1.3.0 · 2026-06-03​

  • Kubernetes cluster monitoring: two modes — static (remote scraping from outside the cluster) and incluster (Prometheus running as a Pod with automatic discovery) — plus new K8s dashboards and menus.
  • A single agent now supports multi-instance middleware monitoring.
  • Loki data retention (30 days by default) and the delete API are enabled.

v1.2.5 · 2026-05-13​

  • The log panel is split into Container Console and Service Logs; Service Logs supports automatic expansion of structured fields and full-text search (phone number / ID / traceID), and all filters are multi-select.
  • Fixed loading issues on middleware panels such as Kafka.
  • Deployment resource requirements corrected to 4C/8G.

v1.2.0 · 2026-04-01​

  • Added log viewing, APM distributed tracing, and a unified interface for metrics / logs / traces.
  • Added support for subpath deployment (reverse-proxy mounted at /mdis, etc.).

v1.1.0 · 2025-02-26​

  • MongoDB slow queries and index suggestions; alert configuration support.

v1.0.0 · 2025-01-02​

  • Initial release: system resource monitoring.