Octri

Monitoring

Search and filters

Every monitoring tab narrows the same way: build a filter visually, or type a query. Both produce the same result, and the query bar is what you'll reach for once you know the fields.

Two ways to filter

Filter builder

Pick a field, an operator, and a value. It only offers operators that make sense for the field's type, so you can't write something invalid.

Query bar

Type an expression directly. Faster once you know the fields, and it's what you paste into an alert.

Fields you can filter on

FieldTypeValues
LevelChoicedebug, info, warning, error, fatal
MethodChoiceGET, POST, PUT, PATCH, DELETE, HEAD, OPTIONS
Status codeNumber
Latency (ms)Number
PathText
OperationText
MessageText
Request IDText
ReleaseText
EnvironmentText
HostText
DeploymentText
AttemptNumber

Operators

The builder offers these by field type. The query bar accepts the same ones.

Field typeOperators
Text=, !=, LIKE, NOT LIKE, IS NULL, IS NOT NULL
Number=, !=, >, >=, <, <=, IS NULL, IS NOT NULL
Choice=, !=, IN, IS NULL, IS NOT NULL

Combine with AND, OR, NOT and parentheses, and finish with ORDER BY <field> [ASC|DESC] or LIMIT <n>.

Examples

sql
level = 'error' AND statusCode >= 500 ORDER BY timestamp DESC
sql
path LIKE '/v1/checkout%' AND latencyMs > 2000
sql
method IN ('POST', 'PATCH') AND release = '9f2c1ab'
LIKE takes % and _ as wildcards

path LIKE '/v1/%' matches every v1 route. % matches any run of characters, _ matches exactly one.

Time range

Filters narrow what; the range narrows when. Pick a preset (1h, 6h, 24h, 7d, 30d, 90d) or set an explicit window.

Beyond 30 days you get the trend, not the events

Raw events are kept 30 days. Charts past that horizon read hourly rollups, so a 90-day view still renders but you cannot open an individual event inside it.

Attempt is the one people forget

Attempt tells you whether an event was a first try or a retry. A wall of errors that are all attempt > 1 is your SDK's retry policy working, not a new outage.

sql
level = 'error' AND attempt = 1

That's the query for real first-time failures, and it is usually a much smaller number.

From a search to an alert

Any query you can run here can back a threshold alert. Get it returning exactly what you'd want to be woken for, then save it as a rule rather than writing it twice.