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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273
// Copyright Materialize, Inc. and contributors. All rights reserved.
//
// Use of this software is governed by the Business Source License
// included in the LICENSE file.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0.
//! A reducible history of compute commands.
use std::borrow::Borrow;
use std::collections::{BTreeMap, BTreeSet};
use mz_ore::cast::CastFrom;
use mz_ore::metrics::UIntGauge;
use mz_ore::{assert_none, soft_assert_or_log};
use timely::progress::Antichain;
use timely::PartialOrder;
use crate::metrics::HistoryMetrics;
use crate::protocol::command::{ComputeCommand, ComputeParameters};
/// TODO(database-issues#7533): Add documentation.
#[derive(Debug)]
pub struct ComputeCommandHistory<M, T = mz_repr::Timestamp> {
/// The number of commands at the last time we compacted the history.
reduced_count: usize,
/// The sequence of commands that should be applied.
///
/// This list may not be "compact" in that there can be commands that could be optimized
/// or removed given the context of other commands, for example compaction commands that
/// can be unified, or dataflows that can be dropped due to allowed compaction.
commands: Vec<ComputeCommand<T>>,
/// Tracked metrics.
metrics: HistoryMetrics<M>,
}
impl<M, T> ComputeCommandHistory<M, T>
where
M: Borrow<UIntGauge>,
T: timely::progress::Timestamp,
{
/// TODO(database-issues#7533): Add documentation.
pub fn new(metrics: HistoryMetrics<M>) -> Self {
metrics.reset();
Self {
reduced_count: 0,
commands: Vec::new(),
metrics,
}
}
/// Add a command to the history.
///
/// This action will reduce the history every time it doubles.
pub fn push(&mut self, command: ComputeCommand<T>) {
self.commands.push(command);
if self.commands.len() > 2 * self.reduced_count {
self.reduce();
} else {
// Refresh reported metrics. `reduce` already refreshes metrics, so we only need to do
// that here in the non-reduce case.
let command = self.commands.last().expect("pushed above");
self.metrics
.command_counts
.for_command(command)
.borrow()
.inc();
if matches!(command, ComputeCommand::CreateDataflow(_)) {
self.metrics.dataflow_count.borrow().inc();
}
}
}
/// Reduces `self.history` to a minimal form.
///
/// This action not only simplifies the issued history, but importantly reduces the instructions
/// to only reference inputs from times that are still certain to be valid. Commands that allow
/// compaction of a collection also remove certainty that the inputs will be available for times
/// not greater or equal to that compaction frontier.
pub fn reduce(&mut self) {
// First determine what the final compacted frontiers will be for each collection.
// These will determine for each collection whether the command that creates it is required,
// and if required what `as_of` frontier should be used for its updated command.
let mut final_frontiers = BTreeMap::new();
let mut created_dataflows = Vec::new();
let mut scheduled_collections = Vec::new();
let mut live_peeks = BTreeMap::new();
let mut create_inst_command = None;
let mut create_timely_command = None;
// Collect only the final configuration.
// Note that this is only correct as long as all config parameters apply globally. If we
// ever introduce parameters that only affect subsequent commands, we will have to
// reconsider this approach.
let mut final_configuration = ComputeParameters::default();
let mut initialization_complete = false;
let mut allow_writes = false;
for command in self.commands.drain(..) {
match command {
create_timely @ ComputeCommand::CreateTimely { .. } => {
assert_none!(create_timely_command);
create_timely_command = Some(create_timely);
}
// We should be able to handle the Create* commands, should this client need to be restartable.
create_inst @ ComputeCommand::CreateInstance(_) => {
assert_none!(create_inst_command);
create_inst_command = Some(create_inst);
}
ComputeCommand::InitializationComplete => {
initialization_complete = true;
}
ComputeCommand::UpdateConfiguration(params) => {
final_configuration.update(params);
}
ComputeCommand::CreateDataflow(dataflow) => {
created_dataflows.push(dataflow);
}
ComputeCommand::Schedule(id) => {
scheduled_collections.push(id);
}
ComputeCommand::AllowCompaction { id, frontier } => {
final_frontiers.insert(id, frontier.clone());
}
ComputeCommand::Peek(peek) => {
live_peeks.insert(peek.uuid, peek);
}
ComputeCommand::CancelPeek { uuid } => {
live_peeks.remove(&uuid);
}
ComputeCommand::AllowWrites => {
allow_writes = true;
}
}
}
// Update dataflow `as_of` frontiers according to allowed compaction.
// One possible frontier is the empty frontier, indicating that the dataflow can be removed.
for dataflow in created_dataflows.iter_mut() {
let mut as_of = Antichain::new();
let initial_as_of = dataflow.as_of.as_ref().unwrap();
for id in dataflow.export_ids() {
// If compaction has been allowed use that; otherwise use the initial `as_of`.
if let Some(frontier) = final_frontiers.get(&id) {
as_of.extend(frontier.clone());
} else {
as_of.extend(initial_as_of.clone());
}
}
soft_assert_or_log!(
PartialOrder::less_equal(initial_as_of, &as_of),
"dataflow as-of regression: {:?} -> {:?} (exports={})",
initial_as_of.elements(),
as_of.elements(),
dataflow.display_export_ids(),
);
// Remove compaction for any collection that brought us to `as_of`.
for id in dataflow.export_ids() {
if let Some(frontier) = final_frontiers.get(&id) {
if frontier == &as_of {
final_frontiers.remove(&id);
}
}
}
dataflow.as_of = Some(as_of);
}
// Discard dataflows whose outputs have all been allowed to compact away.
created_dataflows.retain(|dataflow| dataflow.as_of != Some(Antichain::new()));
let retained_collections: BTreeSet<_> = created_dataflows
.iter()
.flat_map(|d| d.export_ids())
.collect();
scheduled_collections.retain(|id| retained_collections.contains(id));
// Reconstitute the commands as a compact history.
// When we update `metrics`, we need to be careful to not transiently report incorrect
// counts, as they would be observable by other threads.
let command_counts = &self.metrics.command_counts;
let dataflow_count = &self.metrics.dataflow_count;
let count = u64::from(create_timely_command.is_some());
command_counts.create_timely.borrow().set(count);
if let Some(create_timely_command) = create_timely_command {
self.commands.push(create_timely_command);
}
let count = u64::from(create_inst_command.is_some());
command_counts.create_instance.borrow().set(count);
if let Some(create_inst_command) = create_inst_command {
self.commands.push(create_inst_command);
}
let count = u64::from(!final_configuration.all_unset());
command_counts.update_configuration.borrow().set(count);
if !final_configuration.all_unset() {
self.commands
.push(ComputeCommand::UpdateConfiguration(final_configuration));
}
let count = u64::cast_from(created_dataflows.len());
command_counts.create_dataflow.borrow().set(count);
dataflow_count.borrow().set(count);
for dataflow in created_dataflows {
self.commands.push(ComputeCommand::CreateDataflow(dataflow));
}
let count = u64::cast_from(scheduled_collections.len());
command_counts.schedule.borrow().set(count);
for id in scheduled_collections {
self.commands.push(ComputeCommand::Schedule(id));
}
let count = u64::cast_from(live_peeks.len());
command_counts.peek.borrow().set(count);
for peek in live_peeks.into_values() {
self.commands.push(ComputeCommand::Peek(peek));
}
command_counts.cancel_peek.borrow().set(0);
// Allow compaction only after emitting peek commands.
let count = u64::cast_from(final_frontiers.len());
command_counts.allow_compaction.borrow().set(count);
for (id, frontier) in final_frontiers {
self.commands
.push(ComputeCommand::AllowCompaction { id, frontier });
}
let count = u64::from(initialization_complete);
command_counts.initialization_complete.borrow().set(count);
if initialization_complete {
self.commands.push(ComputeCommand::InitializationComplete);
}
if allow_writes {
self.commands.push(ComputeCommand::AllowWrites);
}
self.reduced_count = self.commands.len();
}
/// Discard all peek commands.
pub fn discard_peeks(&mut self) {
self.commands.retain(|command| {
use ComputeCommand::*;
let is_peek = matches!(command, Peek(_) | CancelPeek { .. });
if is_peek {
self.metrics
.command_counts
.for_command(command)
.borrow()
.dec();
}
!is_peek
});
}
/// Iterate through the contained commands.
pub fn iter(&self) -> impl Iterator<Item = &ComputeCommand<T>> {
self.commands.iter()
}
}