Jujing OBSERVE v2.3.0: SQL Search, Alert Templates, and Domestic-Stack Support

New features

SQL log search. The log query page gains a SQL mode supporting WHERE, GROUP BY, JOIN, and aggregate functions, with fields extracted automatically at the collector. Existing users don't need to change collection config — fields are pulled on demand at query time.

Alert rule templates. Save alert rules as templates, parameterize thresholds and targets, and apply one template to dozens of services. Change a threshold in one place and it updates everywhere. Templates support variables (like {service}, {threshold}) and inheritance.

Domestic database support. Metadata and config storage gains adapters for DM8 and KingbaseES V8, with migration scripts from MySQL and a DTS migration guide.

Improvements

• Aggregation is pushed down to the storage layer; COUNT/GROUP BY on a 3-billion-row dataset dropped from 12 s to under 3 s.
• The Collector adds tail-based sampling: errors and slow requests are kept at 100%, normal requests sampled proportionally, cutting trace storage costs by 40% on average.
• Alert notifications add IM bot template variables (WeCom / DingTalk / Feishu), with per-severity channel routing.

Bug fixes

• Fixed occasional missing rows in paginated queries under the DM dialect.
• Fixed span timeline misalignment at time-zone boundaries (UTC / local switching).
• Fixed blank chart rendering in Kylin V10's browser (domestic kernel).
• Fixed duplicate notifications under heavy alert concurrency.

Breaking change

The notify_channels field on alert_rule changes from a comma-separated string to a JSON array. After upgrading, re-save each alert rule or notifications will not fire. The migration script migrates automatically, but export a rule backup before upgrading.

Upgrade steps

# 1. back up database and config
mysqldump -u obs -p observe > backup.sql
# 2. upgrade binaries
systemctl stop observe && \
  tar -xzf observe-2.3.0.tar.gz -C /opt/observe && \
  systemctl start observe
# 3. run migration (automatic)
observe-cli migrate --to 2.3.0
# 4. verify
observe-cli health

After upgrading, run the alert-rule migration in a test environment first and confirm the notify_channels conversion before touching production.

What's next

Planned for v2.4: unified queries across metrics and logs, cross-cluster federated queries, and collection adapters for more domestic middleware (TongWeb, BES). Report issues in the community — high-impact ones get prioritized into the next patch.