be/src/cloud/cloud_throttle_state_machine.h
Line | Count | Source |
1 | | // Licensed to the Apache Software Foundation (ASF) under one |
2 | | // or more contributor license agreements. See the NOTICE file |
3 | | // distributed with this work for additional information |
4 | | // regarding copyright ownership. The ASF licenses this file |
5 | | // to you under the Apache License, Version 2.0 (the |
6 | | // "License"); you may not use this file except in compliance |
7 | | // with the License. You may obtain a copy of the License at |
8 | | // |
9 | | // http://www.apache.org/licenses/LICENSE-2.0 |
10 | | // |
11 | | // Unless required by applicable law or agreed to in writing, |
12 | | // software distributed under the License is distributed on an |
13 | | // "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
14 | | // KIND, either express or implied. See the License for the |
15 | | // specific language governing permissions and limitations |
16 | | // under the License. |
17 | | |
18 | | #pragma once |
19 | | |
20 | | #include <cstdint> |
21 | | #include <map> |
22 | | #include <mutex> |
23 | | #include <string_view> |
24 | | #include <utility> |
25 | | #include <vector> |
26 | | |
27 | | namespace doris::cloud { |
28 | | |
29 | | // ============== Common Types ============== |
30 | | |
31 | | // Load-related RPC types that need table-level QPS statistics |
32 | | enum class LoadRelatedRpc : size_t { |
33 | | PREPARE_ROWSET, |
34 | | COMMIT_ROWSET, |
35 | | UPDATE_TMP_ROWSET, |
36 | | UPDATE_PACKED_FILE_INFO, |
37 | | UPDATE_DELETE_BITMAP, |
38 | | COUNT |
39 | | }; |
40 | | |
41 | | // Get the name string for a LoadRelatedRpc type |
42 | | std::string_view load_related_rpc_name(LoadRelatedRpc rpc); |
43 | | |
44 | | // ============== Data Structures ============== |
45 | | |
46 | | // QPS snapshot: the current QPS of a table on a specific RPC type |
47 | | struct RpcQpsSnapshot { |
48 | | LoadRelatedRpc rpc_type; |
49 | | int64_t table_id; |
50 | | double current_qps; |
51 | | }; |
52 | | |
53 | | // Throttle action: describes what action should be taken |
54 | | struct RpcThrottleAction { |
55 | | enum class Type { SET_LIMIT, REMOVE_LIMIT }; |
56 | | |
57 | | Type type; |
58 | | LoadRelatedRpc rpc_type; |
59 | | int64_t table_id; |
60 | | double qps_limit {0}; // only meaningful for SET_LIMIT |
61 | | // Discard queued reservations when applying a downgraded SET_LIMIT action. |
62 | | bool reset_reservation {false}; |
63 | | }; |
64 | | |
65 | | // ============== ThrottleStateMachine ============== |
66 | | |
67 | | // Parameters for throttle state machine |
68 | | struct RpcThrottleParams { |
69 | | int top_k = 3; // Number of top tables to throttle on each upgrade |
70 | | double ratio = 0.5; // Decay ratio for throttle upgrade |
71 | | double floor_qps = 1.0; // Floor value for table-level QPS limit |
72 | | |
73 | 0 | bool operator==(const RpcThrottleParams& other) const { |
74 | 0 | return top_k == other.top_k && ratio == other.ratio && floor_qps == other.floor_qps; |
75 | 0 | } |
76 | | }; |
77 | | |
78 | | // Pure state machine for throttle upgrade/downgrade decisions |
79 | | // - No time awareness: caller drives events via on_upgrade/on_downgrade |
80 | | // - No config dependency: all parameters passed via constructor/update_params |
81 | | // - No side effects: only returns action descriptions, doesn't touch throttler |
82 | | // - Deterministically testable: same event sequence -> same output |
83 | | class RpcThrottleStateMachine { |
84 | | public: |
85 | | explicit RpcThrottleStateMachine(RpcThrottleParams params); |
86 | | |
87 | | // Runtime update parameters, takes effect on next on_upgrade |
88 | | // Note: existing upgrade history is NOT recalculated |
89 | | void update_params(RpcThrottleParams params); |
90 | | |
91 | | // Process a throttle upgrade event |
92 | | // qps_snapshot: current QPS snapshot for each (rpc, table), provided by caller |
93 | | // Returns: list of actions to execute |
94 | | std::vector<RpcThrottleAction> on_upgrade(const std::vector<RpcQpsSnapshot>& qps_snapshot); |
95 | | |
96 | | // Process a throttle downgrade event (undo the most recent upgrade) |
97 | | // Returns: list of actions to execute |
98 | | std::vector<RpcThrottleAction> on_downgrade(); |
99 | | |
100 | | // Query current state |
101 | | size_t upgrade_level() const; // Current upgrade level |
102 | | double get_current_limit(LoadRelatedRpc rpc_type, int64_t table_id) const; // 0 = no limit |
103 | | RpcThrottleParams get_params() const; |
104 | | |
105 | | private: |
106 | | mutable std::mutex _mtx; |
107 | | |
108 | | RpcThrottleParams _params; |
109 | | |
110 | | // Upgrade history for downgrade rollback |
111 | | // changes: (rpc_type, table_id) -> (old_limit, new_limit) |
112 | | struct UpgradeRecord { |
113 | | std::map<std::pair<LoadRelatedRpc, int64_t>, std::pair<double, double>> changes; |
114 | | }; |
115 | | std::vector<UpgradeRecord> _upgrade_history; |
116 | | |
117 | | // Current active limits for all (rpc, table) |
118 | | std::map<std::pair<LoadRelatedRpc, int64_t>, double> _current_limits; |
119 | | }; |
120 | | |
121 | | // ============== ThrottleCoordinator ============== |
122 | | |
123 | | // Coordinator parameters |
124 | | struct ThrottleCoordinatorParams { |
125 | | // Minimum ticks between upgrades |
126 | | int upgrade_cooldown_ticks = 10; |
127 | | // Ticks after last MS_BUSY to trigger downgrade |
128 | | int downgrade_after_ticks = 60; |
129 | | |
130 | 0 | bool operator==(const ThrottleCoordinatorParams& other) const { |
131 | 0 | return upgrade_cooldown_ticks == other.upgrade_cooldown_ticks && |
132 | 0 | downgrade_after_ticks == other.downgrade_after_ticks; |
133 | 0 | } |
134 | | }; |
135 | | |
136 | | // Pure timing control for upgrade/downgrade triggers |
137 | | // - No time awareness: based on tick count, driven by caller |
138 | | // - No config dependency: all parameters passed via constructor/update_params |
139 | | // |
140 | | // Tick semantics: |
141 | | // - 1 tick = 1 millisecond (fixed unit) |
142 | | // - upgrade_cooldown_ticks and downgrade_after_ticks are in milliseconds |
143 | | // - The tick thread advances time by 1000 ticks (1 second) each iteration |
144 | | class RpcThrottleCoordinator { |
145 | | public: |
146 | | explicit RpcThrottleCoordinator(ThrottleCoordinatorParams params); |
147 | | |
148 | | // Runtime update parameters, takes effect on subsequent report_ms_busy/tick calls |
149 | | // Note: existing tick counts are NOT reset |
150 | | void update_params(ThrottleCoordinatorParams params); |
151 | | |
152 | | // Report a MS_BUSY event |
153 | | // Returns true if upgrade should be triggered |
154 | | bool report_ms_busy(); |
155 | | |
156 | | // Advance by specified number of ticks (caller decides actual time between ticks) |
157 | | // Returns true if downgrade should be triggered |
158 | | bool tick(int64_t ticks = 1); |
159 | | |
160 | | // Tell coordinator whether there are pending upgrades that can be downgraded |
161 | | // Called by the state machine consumer after upgrade/downgrade |
162 | | void set_has_pending_upgrades(bool has); |
163 | | |
164 | | // Query state |
165 | | int64_t ticks_since_last_ms_busy() const; |
166 | | int64_t ticks_since_last_upgrade() const; |
167 | | ThrottleCoordinatorParams get_params() const; |
168 | | |
169 | | private: |
170 | | mutable std::mutex _mtx; |
171 | | |
172 | | ThrottleCoordinatorParams _params; |
173 | | // Counters saturate at their decision thresholds. The MS_BUSY counter is reset |
174 | | // when there is no pending upgrade history, so neither counter grows unbounded. |
175 | | int64_t _ticks_since_last_ms_busy = -1; // -1 means inactive or never received |
176 | | int64_t _ticks_since_last_upgrade = -1; // -1 means never upgraded |
177 | | bool _has_pending_upgrades = false; // Whether there are upgrade records to downgrade |
178 | | }; |
179 | | |
180 | | } // namespace doris::cloud |