prometheus/tsdb/docs/format/head_chunks.md
Fiona Liao fd62dbc291 Update chunk format docs with native histograms and OOO
Signed-off-by: Fiona Liao <fiona.liao@grafana.com>
2024-09-27 18:57:58 +01:00

2.8 KiB

Head Chunks on Disk Format

The following describes the format of a chunks file, which is created in the chunks_head/ directory inside the data directory.

Chunks in the files are referenced from the index by uint64 composed of in-file offset (lower 4 bytes) and segment sequence number (upper 4 bytes).

┌──────────────────────────────┐
│  magic(0x0130BC91) <4 byte>  │
├──────────────────────────────┤
│    version(1) <1 byte>       │
├──────────────────────────────┤
│    padding(0) <3 byte>       │
├──────────────────────────────┤
│ ┌──────────────────────────┐ │
│ │         Chunk 1          │ │
│ ├──────────────────────────┤ │
│ │          ...             │ │
│ ├──────────────────────────┤ │
│ │         Chunk N          │ │
│ └──────────────────────────┘ │
└──────────────────────────────┘

Chunk

Unlike chunks in the on-disk blocks, here we additionally store the series reference that each chunk belongs to and the mint/maxt of the chunks. This is because we don't have an index associated with these chunks, hence this metadata is used while replaying the chunks.

┌─────────────────────┬───────────────────────┬───────────────────────┬───────────────────┬───────────────┬──────────────┬────────────────┐
| series ref <8 byte> | mint <8 byte, uint64> | maxt <8 byte, uint64> | encoding <1 byte> | len <uvarint> | data <bytes> │ CRC32 <4 byte> │
└─────────────────────┴───────────────────────┴───────────────────────┴───────────────────┴───────────────┴──────────────┴────────────────┘

OOO encoding

Head chunks use the highest bit of the encoding field to indicate whether it is out-of-order (1) or not (0). This bit is not set for chunks in the on-disk blocks.