Struct sn_dysfunction::DysfunctionDetection

pub struct DysfunctionDetection {
    pub communication_issues: BTreeMap<XorName, VecDeque<Instant>>,
    pub dkg_issues: BTreeMap<XorName, VecDeque<Instant>>,
    pub probe_issues: BTreeMap<XorName, VecDeque<Instant>>,
    pub knowledge_issues: BTreeMap<XorName, VecDeque<Instant>>,
    pub unfulfilled_ops: BTreeMap<XorName, Vec<(OperationId, Instant)>>,
    /* private fields */
}

Dysfunctional node tracking. Allows various potential issues to be tracked and weighted, with unresposive or suspect nodes being noted on request, against which action can then be taken.

Fields

communication_issues: BTreeMap<XorName, VecDeque<Instant>>

The communication issues logged against a node, along with a timestamp.

dkg_issues: BTreeMap<XorName, VecDeque<Instant>>

The dkg issues logged against a node, along with a timestamp to expire after some time.

probe_issues: BTreeMap<XorName, VecDeque<Instant>>

The probe issues logged against a node and as yet unfulfilled, along with a timestamp to expire after some time.

knowledge_issues: BTreeMap<XorName, VecDeque<Instant>>

The knowledge issues logged against a node, along with a timestamp.

unfulfilled_ops: BTreeMap<XorName, Vec<(OperationId, Instant)>>

The unfulfilled pending request operation issues logged against a node, along with an operation ID.

Implementations

impl DysfunctionDetection

pub fn calculate_scores(&self) -> ScoreResults

Calculate the scores of all nodes being tracked and return them in a node -> score map. There is a map for each type of issue.

The ‘score’ for a node is the number of issues logged against that node, minus the average of the number of issues at all the other nodes.

These scores can then be used to highlight nodes that have a higher score than some particular ratio.

pub fn get_dysfunctional_nodes(&mut self) -> BTreeSet<XorName>

Get a list of nodes whose score is DYSFUNCTION_SCORE_THRESHOLD TODO: order these to act upon most dysfunctional first (the nodes must all ProposeOffline over a dysfunctional node and then immediately vote it off. So any other membershipn changes in flight could block this. thus, we need to be callling this function often until nodes are removed.)

impl DysfunctionDetection

pub fn new(nodes: Vec<XorName>) -> Self

Set up a new Dysfunctional Node Tracker.

pub fn track_issue(&mut self, node_id: XorName, issue_type: IssueType)

Adds an issue to the dysfunction tracker.

The op_id only applies when adding an operational issue.

pub fn request_operation_fulfilled(
    &mut self,
    node_id: &XorName,
    operation_id: OperationId
) -> bool

Removes a pending_operation from the node liveness records. Returns true if a record was removed

pub fn dkg_ack_fulfilled(&mut self, node_id: &XorName)

Removes a DKG session from the node liveness records.

pub fn ae_update_msg_received(&mut self, node_id: &XorName)

Removes a probe tracker from the node liveness records.

pub fn get_unfulfilled_ops(&self, node: XorName) -> Vec<OperationId>

Gets the unfulfilled operation IDs for a given node.

This is for convenience, to wrap reading the concurrent data structure that stores the values.

If there are no unfulfilled operations tracked, an empty list will be returned.

pub fn current_nodes(&self) -> &Vec<XorName>

List all current tracked nodes

pub fn add_new_node(&mut self, node: XorName)

Add a new node to the tracker and recompute closest nodes.

pub fn retain_members_only(&mut self, current_members: BTreeSet<XorName>)

Removes tracked nodes not present in current_members.

Tracked issues related to nodes that were removed will also be removed.

Trait Implementations

impl Clone for DysfunctionDetection

fn clone(&self) -> DysfunctionDetection

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for DysfunctionDetection

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.