rust/neon
1
0
Fork 0
mirror of https://github.com/neondatabase/neon.git synced 2025-12-23 06:09:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat(storage-controller): add node shards api (#8896)

Browse Source 
For control-plane managed tenants, we have the page in the admin console
that lists all tenants on a specific pageserver. But for
storage-controller managed ones, we don't have that functionality for
now.

## Summary of changes

Adds an API that lists all shards on a given node (intention + observed)

---------

Signed-off-by: Alex Chi Z <chi@neon.tech>
This commit is contained in:
Alex Chi Z.
2024-09-07 02:14:21 +08:00
committed by GitHub
parent 30583cb626
commit ac5815b594
7 changed files with 157 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

45
control_plane/storcon_cli/src/main.rs
View File

@@ -4,8 +4,8 @@ use std::{str::FromStr, time::Duration};
use clap::{Parser, Subcommand};
use pageserver_api::{
controller_api::{
NodeAvailabilityWrapper, NodeDescribeResponse, ShardSchedulingPolicy, TenantCreateRequest,
TenantDescribeResponse, TenantPolicyRequest,
NodeAvailabilityWrapper, NodeDescribeResponse, NodeShardResponse, ShardSchedulingPolicy,
TenantCreateRequest, TenantDescribeResponse, TenantPolicyRequest,
},
models::{
EvictionPolicy, EvictionPolicyLayerAccessThreshold, LocationConfigSecondary,
@@ -80,7 +80,10 @@ enum Command {
/// List nodes known to the storage controller
Nodes {},
/// List tenants known to the storage controller
Tenants {},
Tenants {
/// If this field is set, it will list the tenants on a specific node
node_id: Option<NodeId>,
},
/// Create a new tenant in the storage controller, and by extension on pageservers.
TenantCreate {
#[arg(long)]
@@ -403,7 +406,41 @@ async fn main() -> anyhow::Result<()> {
)
.await?;
}
Command::Tenants {} => {
Command::Tenants {
node_id: Some(node_id),
} => {
let describe_response = storcon_client
.dispatch::<(), NodeShardResponse>(
Method::GET,
format!("control/v1/node/{node_id}/shards"),
None,
)
.await?;
let shards = describe_response.shards;
let mut table = comfy_table::Table::new();
table.set_header([
"Shard",
"Intended Primary/Secondary",
"Observed Primary/Secondary",
]);
for shard in shards {
table.add_row([
format!("{}", shard.tenant_shard_id),
match shard.is_intended_secondary {
None => "".to_string(),
Some(true) => "Secondary".to_string(),
Some(false) => "Primary".to_string(),
},
match shard.is_observed_secondary {
None => "".to_string(),
Some(true) => "Secondary".to_string(),
Some(false) => "Primary".to_string(),
},
]);
}
println!("{table}");
}
Command::Tenants { node_id: None } => {
let mut resp = storcon_client
.dispatch::<(), Vec<TenantDescribeResponse>>(
Method::GET,

15
libs/pageserver_api/src/controller_api.rs
View File

@@ -112,6 +112,21 @@ pub struct TenantDescribeResponse {
pub config: TenantConfig,
}
#[derive(Serialize, Deserialize, Debug)]
pub struct NodeShardResponse {
pub node_id: NodeId,
pub shards: Vec<NodeShard>,
}
#[derive(Serialize, Deserialize, Debug)]
pub struct NodeShard {
pub tenant_shard_id: TenantShardId,
/// Whether the shard is observed secondary on a specific node. True = yes, False = no, None = not on this node.
pub is_observed_secondary: Option<bool>,
/// Whether the shard is intended to be a secondary on a specific node. True = yes, False = no, None = not on this node.
pub is_intended_secondary: Option<bool>,
}
#[derive(Serialize, Deserialize)]
pub struct NodeDescribeResponse {
pub id: NodeId,

18
storage_controller/src/http.rs
View File

@@ -539,6 +539,17 @@ async fn handle_node_status(req: Request<Body>) -> Result<Response<Body>, ApiErr
json_response(StatusCode::OK, node_status)
}
async fn handle_node_shards(req: Request<Body>) -> Result<Response<Body>, ApiError> {
check_permissions(&req, Scope::Admin)?;
let state = get_state(&req);
let node_id: NodeId = parse_request_param(&req, "node_id")?;
let node_status = state.service.get_node_shards(node_id).await?;
json_response(StatusCode::OK, node_status)
}
async fn handle_get_leader(req: Request<Body>) -> Result<Response<Body>, ApiError> {
check_permissions(&req, Scope::Admin)?;
@@ -1109,6 +1120,13 @@ pub fn make_router(
.get("/control/v1/node/:node_id", |r| {
named_request_span(r, handle_node_status, RequestName("control_v1_node_status"))
})
.get("/control/v1/node/:node_id/shards", |r| {
named_request_span(
r,
handle_node_shards,
RequestName("control_v1_node_describe"),
)
})
.get("/control/v1/leader", |r| {
named_request_span(r, handle_get_leader, RequestName("control_v1_get_leader"))
})

49
storage_controller/src/service.rs
View File

@@ -41,11 +41,11 @@ use itertools::Itertools;
use pageserver_api::{
controller_api::{
MetadataHealthRecord, MetadataHealthUpdateRequest, NodeAvailability, NodeRegisterRequest,
NodeSchedulingPolicy, PlacementPolicy, ShardSchedulingPolicy, ShardsPreferredAzsRequest,
ShardsPreferredAzsResponse, TenantCreateRequest, TenantCreateResponse,
TenantCreateResponseShard, TenantDescribeResponse, TenantDescribeResponseShard,
TenantLocateResponse, TenantPolicyRequest, TenantShardMigrateRequest,
TenantShardMigrateResponse,
NodeSchedulingPolicy, NodeShard, NodeShardResponse, PlacementPolicy, ShardSchedulingPolicy,
ShardsPreferredAzsRequest, ShardsPreferredAzsResponse, TenantCreateRequest,
TenantCreateResponse, TenantCreateResponseShard, TenantDescribeResponse,
TenantDescribeResponseShard, TenantLocateResponse, TenantPolicyRequest,
TenantShardMigrateRequest, TenantShardMigrateResponse,
},
models::{
SecondaryProgress, TenantConfigRequest, TimelineArchivalConfigRequest,
@@ -4924,6 +4924,45 @@ impl Service {
))
}
pub(crate) async fn get_node_shards(
&self,
node_id: NodeId,
) -> Result<NodeShardResponse, ApiError> {
let locked = self.inner.read().unwrap();
let mut shards = Vec::new();
for (tid, tenant) in locked.tenants.iter() {
let is_intended_secondary = match (
tenant.intent.get_attached() == &Some(node_id),
tenant.intent.get_secondary().contains(&node_id),
) {
(true, true) => {
return Err(ApiError::InternalServerError(anyhow::anyhow!(
"{} attached as primary+secondary on the same node",
tid
)))
}
(true, false) => Some(false),
(false, true) => Some(true),
(false, false) => None,
};
let is_observed_secondary = if let Some(ObservedStateLocation { conf: Some(conf) }) =
tenant.observed.locations.get(&node_id)
{
Some(conf.secondary_conf.is_some())
} else {
None
};
if is_intended_secondary.is_some() || is_observed_secondary.is_some() {
shards.push(NodeShard {
tenant_shard_id: *tid,
is_intended_secondary,
is_observed_secondary,
});
}
}
Ok(NodeShardResponse { node_id, shards })
}
pub(crate) async fn get_leader(&self) -> DatabaseResult<Option<ControllerPersistence>> {
self.persistence.get_leader().await
}

8
test_runner/fixtures/common_types.py
View File

@@ -140,6 +140,14 @@ class TenantId(Id):
return self.id.hex()
class NodeId(Id):
def __repr__(self) -> str:
return f'`NodeId("{self.id.hex()}")'
def __str__(self) -> str:
return self.id.hex()
class TimelineId(Id):
def __repr__(self) -> str:
return f'TimelineId("{self.id.hex()}")'

26
test_runner/fixtures/neon_fixtures.py
View File

@@ -62,7 +62,7 @@ from urllib3.util.retry import Retry
from fixtures import overlayfs
from fixtures.broker import NeonBroker
from fixtures.common_types import Lsn, TenantId, TenantShardId, TimelineId
from fixtures.common_types import Lsn, NodeId, TenantId, TenantShardId, TimelineId
from fixtures.endpoint.http import EndpointHttpClient
from fixtures.log_helper import log
from fixtures.metrics import Metrics, MetricsGetter, parse_metrics
@@ -2570,6 +2570,30 @@ class NeonStorageController(MetricsGetter, LogUtils):
response.raise_for_status()
return response.json()
def nodes(self):
"""
:return: list of {"id": ""}
"""
response = self.request(
"GET",
f"{self.api}/control/v1/node",
headers=self.headers(TokenScope.ADMIN),
)
response.raise_for_status()
return response.json()
def node_shards(self, node_id: NodeId):
"""
:return: list of {"shard_id": "", "is_secondary": bool}
"""
response = self.request(
"GET",
f"{self.api}/control/v1/node/{node_id}/shards",
headers=self.headers(TokenScope.ADMIN),
)
response.raise_for_status()
return response.json()
def tenant_shard_split(
self, tenant_id: TenantId, shard_count: int, shard_stripe_size: Optional[int] = None
) -> list[TenantShardId]:

6
test_runner/regress/test_storage_controller.py
View File

@@ -1552,6 +1552,12 @@ def test_tenant_import(neon_env_builder: NeonEnvBuilder, shard_count, remote_sto
literal_shard_count = 1 if shard_count is None else shard_count
assert len(describe["shards"]) == literal_shard_count
nodes = env.storage_controller.nodes()
assert len(nodes) == 2
describe1 = env.storage_controller.node_shards(nodes[0]["id"])
describe2 = env.storage_controller.node_shards(nodes[1]["id"])
assert len(describe1["shards"]) + len(describe2["shards"]) == literal_shard_count
# Check the data is still there: this implicitly proves that we recovered generation numbers
# properly, for the timeline which was written to after a generation bump.
for timeline, branch, expect_rows in [