2018-05-30 08:05:30 -07:00
# WAL Disk Format
2018-07-19 23:26:12 -07:00
The write ahead log operates in segments that are numbered and sequential,
2018-05-30 08:05:30 -07:00
e.g. `000000` , `000001` , `000002` , etc., and are limited to 128MB by default.
A segment is written to in pages of 32KB. Only the last page of the most recent segment
may be partial. A WAL record is an opaque byte slice that gets split up into sub-records
should it exceed the remaining space of the current page. Records are never split across
2018-06-18 04:52:57 -07:00
segment boundaries. If a single record exceeds the default segment size, a segment with
a larger size will be created.
2018-08-02 14:46:45 -07:00
The encoding of pages is largely borrowed from [LevelDB's/RocksDB's write ahead log. ](https://github.com/facebook/rocksdb/wiki/Write-Ahead-Log-File-Format )
2018-05-30 08:05:30 -07:00
Notable deviations are that the record fragment is encoded as:
2018-06-18 04:52:57 -07:00
```
2018-05-30 08:05:30 -07:00
┌───────────┬──────────┬────────────┬──────────────┐
│ type < 1b > │ len < 2b > │ CRC32 < 4b > │ data < bytes > │
└───────────┴──────────┴────────────┴──────────────┘
2018-06-18 04:52:57 -07:00
```
2018-05-30 08:05:30 -07:00
2018-07-19 23:26:12 -07:00
The type flag has the following states:
* `0` : rest of page will be empty
* `1` : a full record encoded in a single fragment
* `2` : first fragment of a record
* `3` : middle fragment of a record
* `4` : final fragment of a record
2018-05-30 08:05:30 -07:00
## Record encoding
The records written to the write ahead log are encoded as follows:
### Series records
2018-07-19 23:26:12 -07:00
Series records encode the labels that identifies a series and its unique ID.
2018-05-30 08:05:30 -07:00
2018-06-18 04:52:57 -07:00
```
2018-05-30 08:05:30 -07:00
┌────────────────────────────────────────────┐
│ type = 1 < 1b > │
├────────────────────────────────────────────┤
│ ┌─────────┬──────────────────────────────┐ │
│ │ id < 8b > │ n = len(labels) < uvarint > │ │
│ ├─────────┴────────────┬─────────────────┤ │
│ │ len(str_1) < uvarint > │ str_1 < bytes > │ │
│ ├──────────────────────┴─────────────────┤ │
│ │ ... │ │
│ ├───────────────────────┬────────────────┤ │
│ │ len(str_2n) < uvarint > │ str_2n < bytes > │ │
│ └───────────────────────┴────────────────┘ │
│ . . . │
└────────────────────────────────────────────┘
2018-06-18 04:52:57 -07:00
```
2018-05-30 08:05:30 -07:00
### Sample records
Sample records encode samples as a list of triples `(series_id, timestamp, value)` .
Series reference and timestamp are encoded as deltas w.r.t the first sample.
2019-03-21 02:21:50 -07:00
The first row stores the starting id and the starting timestamp.
The first sample record begins at the second row.
2018-05-30 08:05:30 -07:00
2018-06-18 04:52:57 -07:00
```
2018-05-30 08:05:30 -07:00
┌──────────────────────────────────────────────────────────────────┐
│ type = 2 < 1b > │
├──────────────────────────────────────────────────────────────────┤
2019-03-21 02:21:50 -07:00
│ ┌────────────────────┬───────────────────────────┐ │
│ │ id < 8b > │ timestamp < 8b > │ │
│ └────────────────────┴───────────────────────────┘ │
2018-05-30 08:05:30 -07:00
│ ┌────────────────────┬───────────────────────────┬─────────────┐ │
│ │ id_delta < uvarint > │ timestamp_delta < uvarint > │ value < 8b > │ │
│ └────────────────────┴───────────────────────────┴─────────────┘ │
│ . . . │
└──────────────────────────────────────────────────────────────────┘
2018-06-18 04:52:57 -07:00
```
2018-05-30 08:05:30 -07:00
### Tombstone records
Tombstone records encode tombstones as a list of triples `(series_id, min_time, max_time)`
and specify an interval for which samples of a series got deleted.
2018-06-18 04:52:57 -07:00
```
2018-05-30 08:05:30 -07:00
┌─────────────────────────────────────────────────────┐
│ type = 3 < 1b > │
├─────────────────────────────────────────────────────┤
│ ┌─────────┬───────────────────┬───────────────────┐ │
│ │ id < 8b > │ min_time < varint > │ max_time < varint > │ │
│ └─────────┴───────────────────┴───────────────────┘ │
│ . . . │
└─────────────────────────────────────────────────────┘
2018-06-18 04:52:57 -07:00
```
2021-05-06 13:53:52 -07:00
### Exemplar records
Exemplar records encode exemplars as a list of triples `(series_id, timestamp, value)`
plus the length of the labels list, and all the labels.
The first row stores the starting id and the starting timestamp.
Series reference and timestamp are encoded as deltas w.r.t the first exemplar.
The first exemplar record begins at the second row.
See: https://github.com/OpenObservability/OpenMetrics/blob/main/specification/OpenMetrics.md#exemplars
```
┌──────────────────────────────────────────────────────────────────┐
│ type = 5 < 1b > │
├──────────────────────────────────────────────────────────────────┤
│ ┌────────────────────┬───────────────────────────┐ │
│ │ id < 8b > │ timestamp < 8b > │ │
│ └────────────────────┴───────────────────────────┘ │
│ ┌────────────────────┬───────────────────────────┬─────────────┐ │
│ │ id_delta < uvarint > │ timestamp_delta < uvarint > │ value < 8b > │ │
│ ├────────────────────┴───────────────────────────┴─────────────┤ │
│ │ n = len(labels) < uvarint > │ │
│ ├──────────────────────┬───────────────────────────────────────┤ │
│ │ len(str_1) < uvarint > │ str_1 < bytes > │ │
│ ├──────────────────────┴───────────────────────────────────────┤ │
│ │ ... │ │
│ ├───────────────────────┬──────────────────────────────────────┤ │
│ │ len(str_2n) < uvarint > │ str_2n < bytes > │ │ │
│ └───────────────────────┴────────────────┴─────────────────────┘ │
│ . . . │
└──────────────────────────────────────────────────────────────────┘
```