Query Traces
Query operation queries the data in a trace.
bydbctl is the command line tool in examples.
The input contains two parts:
- Request: a YAML-based text which is defined by the API
- Time Range: YAML and CLI’s flags both support it.
Time Range
The query specification contains time_range field. The request should set absolute times to it.
bydbctl also provides start and end flags to support passing absolute and relative times.
“start” and “end” specify a time range during which the query is performed, they can be an absolute time like “2006-01-02T15:04:05Z07:00”, or relative time (to the current time) like “-30m”, or “30m”. They are both optional and their default values follow the rules below:
- when “start” and “end” are both absent, “start = now - 30 minutes” and “end = now”, namely past 30 minutes;
- when “start” is absent and “end” is present, this command calculates “start” (minus 30 units), e.g. “end = 2022-11-09T12:34:00Z”, so “start = end - 30 minutes = 2022-11-09T12:04:00Z”;
- when “start” is present and “end” is absent, this command calculates “end” (plus 30 units), e.g. “start = 2022-11-09T12:04:00Z”, so “end = start + 30 minutes = 2022-11-09T12:34:00Z”.
Understand the schema you are querying
Before querying the data, you need to know the trace name and the tags defined in the trace. You can use the bydbctl trace get command to get the trace schema.
For example, if you want to get the schema of a trace named sw in the group sw_trace, you can use the below command:
bydbctl trace get -g sw_trace -n sw
trace:
metadata:
createRevision: "100"
group: sw_trace
id: 0
modRevision: "100"
name: sw
tags:
- name: trace_id
type: TAG_TYPE_STRING
- name: state
type: TAG_TYPE_INT
- name: service_id
type: TAG_TYPE_STRING
- name: service_instance_id
type: TAG_TYPE_STRING
- name: endpoint_id
type: TAG_TYPE_STRING
- name: duration
type: TAG_TYPE_INT
- name: span_id
type: TAG_TYPE_STRING
- name: timestamp
type: TAG_TYPE_TIMESTAMP
traceIdTagName: trace_id
spanIdTagName: span_id
timestampTagName: timestamp
updatedAt: null
Examples
The following examples use the above schema to show how to query data in a trace and cover some common use cases:
Query between specific time range
To retrieve traces named sw between 2022-10-15T22:32:48Z and 2022-10-15T23:32:48Z, the response will contain matching spans grouped by trace ID. These traces also choose a tag projection trace_id.
bydbctl trace query -f - <<EOF
groups: ["sw_trace"]
name: "sw"
tag_projection: ["trace_id"]
timeRange:
begin: 2022-10-15T22:32:48+08:00
end: 2022-10-15T23:32:48+08:00
EOF
Query using relative time duration
The below command could query data in the last 30 minutes using relative time duration:
bydbctl trace query --start -30m -f - <<EOF
groups: ["sw_trace"]
name: "sw"
tag_projection: ["trace_id"]
EOF
Query with filter by trace ID
The below command could query data filtered by a specific trace ID:
bydbctl trace query -f - <<EOF
name: "sw"
groups: ["sw_trace"]
tag_projection: ["trace_id"]
criteria:
condition:
name: "trace_id"
op: "BINARY_OP_EQ"
value:
str:
value: "trace_001"
EOF
Query with filter
The below command could query data with multiple filter conditions using logical AND:
bydbctl trace query -f - <<EOF
name: "sw"
groups: ["sw_trace"]
criteria:
le:
op: "LOGICAL_OP_AND"
left:
condition:
name: "service_instance_id"
op: "BINARY_OP_EQ"
value:
str:
value: "webapp_instance_1"
right:
condition:
name: "endpoint_id"
op: "BINARY_OP_EQ"
value:
str:
value: "/home_endpoint"
orderBy:
indexRuleName: "timestamp"
sort: SORT_ASC
EOF
More filter operations can be found in here.
Query ordered by index
The below command could query data ordered by duration in descending order:
bydbctl trace query -f - <<EOF
name: "sw"
groups: ["sw_trace"]
orderBy:
indexRuleName: "duration"
sort: "SORT_DESC"
EOF
Query with limit
The below command could query ordered data and return the first two results:
bydbctl trace query -f - <<EOF
name: "sw"
groups: ["sw_trace"]
orderBy:
indexRuleName: "timestamp"
sort: "SORT_DESC"
limit: 2
EOF
Query from Multiple Groups
When querying data from multiple groups, you can combine traces that share the same trace name. Note the following requirements:
- Tags with the same name across different groups must share the same type.
- The orderBy indexRuleName must be identical across groups.
- Any tags used in filter criteria must exist in all groups with the same name and type.
bydbctl trace query -f - <<EOF
name: "sw"
groups: ["sw_trace", "another_trace_group"]
tag_projection: ["trace_id", "service_id"]
criteria:
condition:
name: "trace_id"
op: "BINARY_OP_IN"
value:
str_array:
value: ["trace_001", "trace_002"]
EOF