linkrs-storage 0.1.0

Storage engine with CSR and memory-mapped containers
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190

use crate::engine::resource_budget::{MemoryCategory, ResourceSnapshot};

use super::GraphStorageContext;

impl GraphStorageContext {
    pub(crate) fn storage_size(&self) -> usize {
        let vertices = self.persistent.data_store.with_vertex_tables(|tables| {
            tables
                .values()
                .map(|table| table.memory_size())
                .sum::<usize>()
        });
        let edges = self.persistent.data_store.with_edge_tables(|tables| {
            tables
                .values()
                .map(|arc| arc.read().memory_size())
                .sum::<usize>()
        });
        vertices + edges
    }

    pub(crate) fn used_storage_size(&self) -> usize {
        let vertices = self.persistent.data_store.with_vertex_tables(|tables| {
            tables
                .values()
                .map(|table| table.used_memory_size())
                .sum::<usize>()
        });
        let edges = self.persistent.data_store.with_edge_tables(|tables| {
            tables
                .values()
                .map(|arc| arc.read().used_memory_size())
                .sum::<usize>()
        });
        vertices + edges
    }

    pub fn resource_snapshot(&self) -> ResourceSnapshot {
        self.persistent
            .resource_accounting
            .report_usage(MemoryCategory::Data, self.used_storage_size() as u64);
        let index_bytes = self
            .persistent
            .index_data_manager
            .read()
            .cached_memory_usage_bytes();
        self.persistent
            .resource_accounting
            .report_usage(MemoryCategory::Index, index_bytes);
        let tombstone_count = self
            .persistent
            .index_data_manager
            .read()
            .cached_tombstone_count() as usize;
        let tombstone_memory_bytes = (tombstone_count as u64).saturating_mul(64);
        self.persistent
            .resource_accounting
            .report_usage(MemoryCategory::Mvcc, tombstone_memory_bytes);
        let _ = self.persistent.cache_manager.refresh_memory_usage();
        // Periodic export point: push collected bloom pre-check counters to
        // the StatsManager alongside the other resource metrics.
        self.persistent
            .index_data_manager
            .read()
            .export_bloom_stats();
        let mut snapshot = self.persistent.resource_accounting.snapshot();
        snapshot.active_snapshots = self
            .persistent
            .version_manager
            .snapshot_tracker()
            .active_count();
        snapshot.oldest_snapshot_ts = self
            .persistent
            .version_manager
            .snapshot_tracker()
            .cleanup_threshold();
        snapshot.tombstone_count = tombstone_count;
        snapshot.tombstone_memory_bytes = tombstone_memory_bytes;
        snapshot.vertex_shard_count = self
            .persistent
            .data_store
            .with_vertex_tables(|tables| tables.values().map(|table| table.num_shards()).sum());
        // Keep spiller accessors exercised.
        let _spill_ratio = self.spiller().spill_threshold_ratio();
        let _spill_dir = self.spiller().spill_dir();
        let _active_spills = self.spiller().active_spills().read().len();
        // Exercise try_reserve_with_spill with a zero-byte probe to keep
        // the full reservation-with-spill path compiled and tested.
        let _probe = self.try_reserve_with_spill(MemoryCategory::Data, 0);
        // Keep vertex GC stats exercised.
        if let Some(ref gc) = self.runtime.vertex_gc_manager {
            let _total = gc.total_removed();
            let _passes = gc.pass_count();
            let _mitigations = gc.pressure_mitigations();
        }
        snapshot
    }

    pub fn check_write_admission(&self) -> linkrs_core::StorageResult<()> {
        if self
            .operation_context
            .as_ref()
            .is_some_and(|context| context.read_only)
        {
            return Err(linkrs_core::StorageError::invalid_operation(
                "Read-only transaction cannot perform writes",
            ));
        }
        let snapshot = self.resource_snapshot();
        if snapshot.hard_limit_exceeded() {
            // Before failing, attempt to spill cold data to recover memory.
            let overage = snapshot
                .total_current_bytes
                .saturating_sub(snapshot.budget.max_memory_bytes)
                + 1024 * 1024;
            self.spiller().spill_cold_data(overage);
            let snapshot = self.resource_snapshot();
            if snapshot.hard_limit_exceeded() {
                return Err(linkrs_core::StorageError::capacity_exceeded());
            }
        }
        let resources = &self.persistent.config.resources;
        if snapshot.tombstone_count >= resources.max_tombstones
            || snapshot.tombstone_memory_bytes >= resources.max_tombstone_bytes
        {
            return Err(linkrs_core::StorageError::capacity_exceeded());
        }
        if snapshot.soft_limit_exceeded() {
            log::debug!(
                "Storage memory is above the soft limit: {} / {} bytes",
                snapshot.total_current_bytes,
                snapshot.budget.max_memory_bytes
            );
        }
        // Version-chain memory diagnostic and long-transaction backpressure.
        // Long-lived snapshots pin history; we warn and surface
        // blocked bytes but do not force the safe GC frontier forward.
        {
            let coordinator = crate::engine::gc_coordinator::GcCoordinator::new(
                self.persistent.version_manager.clone(),
            );
            let diag = coordinator.diagnostics();
            if diag.is_blocked() {
                let version_bytes: usize =
                    self.persistent.data_store.with_vertex_tables(|tables| {
                        tables
                            .values()
                            .map(|t| t.version_chain_memory_bytes())
                            .sum()
                    });
                if version_bytes > 64 * 1024 * 1024 {
                    log::warn!(
                        "write admission: long transaction blocks GC ({} bytes version chains, {} snapshots)",
                        version_bytes,
                        diag.active_snapshot_count
                    );
                }
            }
        }
        Ok(())
    }

    pub fn check_snapshot_admission(&self) -> linkrs_core::StorageResult<()> {
        let tracker = self.persistent.version_manager.snapshot_tracker();
        let active = tracker.active_count();
        if active >= self.persistent.config.resources.max_active_snapshots {
            return Err(linkrs_core::StorageError::capacity_exceeded());
        }
        if let Some(age) = tracker.oldest_age() {
            if age >= self.persistent.config.resources.max_snapshot_age {
                return Err(linkrs_core::StorageError::invalid_operation(
                    "Oldest active snapshot exceeded max_snapshot_age",
                ));
            }
            // Long-transaction diagnostic: warn when oldest snapshot
            // ages past 30s and blocks GC. Does not force watermark forward;
            // caller may retry or apply backpressure upstream.
            if age.as_secs() > 30 {
                let coordinator = crate::engine::gc_coordinator::GcCoordinator::new(
                    self.persistent.version_manager.clone(),
                );
                let diag = coordinator.diagnostics();
                if let Some(warn) = diag.long_transaction_warning {
                    log::warn!("snapshot admission blocked by long transaction: {}", warn);
                }
            }
        }
        Ok(())
    }
}