opsx
7369a17598
feat(alerts): 数据库连接告警覆盖 MySQL/PG/Redis 主流报错;规则表单三字段对齐
=== 1. 「数据库连接失败」匹配模式 3 → 25 个 ===
原来只匹配 connection refused / reset by peer / too many connections,
漏掉了 PG 的 FATAL: too many clients、Redis 的 NOAUTH / WRONGPASS /
connection pool timeout 等实际最常见的报错。
按数据库分类补齐:
通用网络层
connection refused / connection reset by peer / broken pipe /
no route to host / cannot assign requested address /
connection timed out / i/o timeout / unexpected EOF /
server closed the connection unexpectedly
MySQL
too many connections / communications link failure /
driver: bad connection / could not create connection /
Access denied for user
PostgreSQL
FATAL: password authentication failed / sorry, too many clients /
terminating connection
Redis
NOAUTH / WRONGPASS / READONLY you can't write against /
CLUSTERDOWN / max number of clients reached
连接池(Java/Go 应用层最先炸的地方)
connection pool / connection is not available
**刻意不加**的(会匹配到几乎所有网络日志,等于没告警):
裸 "EOF"、"timeout"、"error"、"refused"、"deadlock"
阈值保持 1000/15min(模式变多不代表命中变多,都是真错误才命中)。
迁移 v9→v10 只替换"仍是旧默认查询"的规则,用户自定义过的不动。
=== 2. 规则表单三个字段没对齐 ===
现象:统计窗口(分) / 触发阈值(条) / 冷却(分) 三格并排,
但只有中间的「触发阈值」下方多了一行提示文字,
导致中间那格更高、提示文字悬在底部,看起来像没对齐。
修法:三格统一为「标签 + 输入框 + 一行提示(min-h-4 固定高度)」,
底边自然齐平;提示文案改为:
统计窗口 → 统计最近多少分钟
触发阈值 → 超过该条数才告警
冷却 → 冷却内不重复发送
(顺带把阈值提示从"0 = 出现即告警"改掉:默认阈值已无 0,0 反而是危险配置)
小屏改单列堆叠(grid-cols-1 sm:grid-cols-3)。
=== 实测(936a127b1d15)===
v10 迁移:告警规则 "数据库连接失败" 扩展匹配模式
线上规则:25 个模式,含 NOAUTH / too many clients,阈值 1000,冷却 240m
Pod = latest = v0.1 = 936a127b1d15