drand supports multiple storage backends: BoltDB, PostgreSQL, in-memory storage, aka. MemDB.
To launch drand with a specific backend, you need to provide the --db
flag.
The supported values are: bolt
for BoltDB, postgres
for PostgreSQL, memdb
for the in-memory backend.
BoltDB
The BoltDB backend allows drand to store beacon data under a local filesystem.
The database can be found under the <storage-path>/multibeacon/<beacon-id>/db/drand.db
.
This is the default storage backend.
Usage Example:
drand start \
-db boltdb \
<other "start" options here>
You can also omit the -db
flag.
At the moment, there are no configurable options for this backend.
🐘 PostgreSQL
This backend is supported since drand v1.5+.
The PostgreSQL backend allows drand to store beacons in a database.
The database connection will use UTC
as default timezone. This parameter is not currently configurable.
Usage Example:
Before using this backend, you'll need to create a user that can read/write into the database you want the beacons to be stored in.
You'll also need to manually apply the migration schema chain/postgresdb/schema/schema.sql.
Note: You should always apply the schema corresponding to your code version.
drand start \
-db postgres \
-pg-dsn 'postgres://drand:[email protected]:45432/drand?sslmode=disable&connect_timeout=5' \
<other "start" options here>
Supported Options:
You can customize the PostgreSQL backend using the -pg-dsn
option. This allows you to control various connection properties.
The available options are:
sslmode
- If the SSL mode is disabled or not. Supports:required
ordisabled
. Defaultdisabled
.connect_timeout
- How many seconds to wait until the connection attempt times out. Default: 5 seconds.max-idle
- The number of maximum idle connections to the database. Default: 2.max-open
- The number of maximum open connections to the database. Default: 0 - unlimited.- If
max-idle
is greater than 0 andmax-open
is less thanmax-idle
, thenmax-idle
will be capped tomax-open
limit.
In-memory (memDB)
This backend is supported since drand v1.5+.
The in-memory backend, also referred to as memdb, allows the node to store all beacons in-memory.
This backend will lose all the contents on node restart/stop.
Warning: Using this backend on a threshold amount of nodes can cause catastrophic network disruptions if all nodes reboot/fail at the same time.
Usage Example:
drand start \
-db memdb \
-memdb-size 2000 \
<other "start" options here>
Supported Options:
You can customize the in-memory backend by specifying the following options:
memdb-size
- The number of beacons that each network will store before purging.
Note, the memdb-size
applies to individually to all the networks. E.g., for a -memdb-size=2000
:
- If you plan to run a single network, then the total number of beacons in memory will be 2000.
- If you plan to run two networks, then the total number of beacons in memory will be 4000.
- If you plan to run N networks, then the total number of beacons in memory will be N * 2000.
To help you estimate the memory size required to operate an in-memory node, you can use a size of about 200 bytes for each beacon value stored.
Besides the memory constraints, you'll also have to keep in mind the frequency of the network:
- For a 30 seconds period, 2000 beacons will cover the last ~16 hours of values.
- For a 3 seconds period, 2000 beacons will cover the last ~1 hour of values.
- For a 1 second period, 2000 beacons will cover the last ~30 minutes of values