1. Second line
Each newline is another point; one request can carry thousands of lines.
cpu,host=b usage=0.71 1700000000Typorithm · medium
Batching Writes
write-batchinginfluxdbbatchingcardinalitydesign
Saved on this device
Full lesson · 8 learning steps Open any exercise for focused practice
10%
Step 1 of 8 · intro card
Learn Batching Writes
Line Protocol
Batch many lines into one request for throughput, model low-cardinality identifiers as indexed tags and raw values as fields, and pick a timestamp precision that fits the whole batch.
Sending points one at a time is slow; batching many newline-separated lines per request cuts round trips dramatically.
Design each line deliberately: tags are indexed and should stay low-cardinality (host, region), while fields hold the unindexed measured values.
High-cardinality data as tags (like a user id) explodes the index, so push those into fields and share one --precision across the batch.
01
Many newline-separated lines = one efficient request.
02
Tags are indexed; keep them low-cardinality.
03
Fields are unindexed values; high-cardinality data belongs here.
influx write -b metrics --precision s "cpu,host=a usage=0.64 1700000000
cpu,host=b usage=0.71 1700000000
mem,host=a used=812 1700000000"Read the implementation
Connect each part of the code to the concept before you type.
2. Mixed measurements
A single batch may mix measurements and tags freely.
mem,host=a used=812 17000000003. Indexed tag
host is a tag — indexed and low-cardinality, ideal for filtering and grouping.
host=a4. Shared precision
One precision flag governs every timestamp in the batch.
--precision sA complete learning loop
01
Understand
Read the concept map and follow the code breakdown.
02
Reconstruct
Type the implementation, repair mistakes and work through every recall exercise.
03
Verify
Use the final check without hints; revisit weak concepts in short practice sessions.