Closed Super-long closed 2 weeks ago
Maybe we can add a name field to the object store config. The ObjectStoreManager
uses names to distinguish storages.
https://github.com/GreptimeTeam/greptimedb/blob/2d992f4f12c70aaa22477208aee05aa5ee8979e5/src/datanode/src/config.rs#L49-L59
Maybe we can add a name field to the object store config. The
ObjectStoreManager
uses names to distinguish storages.
yes, that's my plan. Adds a name
for storage configs.
@Super-long Already support it by name
in storage.providers
, for example
[[storage.providers]]
name = "test1"
type = "S3"
bucket = "test_greptimedb"
root = "/greptimedb"
access_key_id = "<access key id>"
secret_access_key = "<secret access key>"
Then you can create the table with this provider name:
CREATE TABLE IF NOT EXISTS temperatures(
ts TIMESTAMP TIME INDEX,
temperature DOUBLE DEFAULT 10,
) engine=mito with(ttl='7d', storage="test1");
This feature will be shipped in the v0.9.3 release.
What problem does the new feature solve?
Enables single cluster throughput to be horizontally scalable
What does the feature do?
Supports multiple buckets for object storage in a single cluster
Implementation challenges
No response