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
// 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.
//! All machines need maintenance
//!
//! Maintenance operations for persist, shared among active handles
use std::fmt::Debug;
use std::mem;
use std::sync::Arc;
use differential_dataflow::difference::Semigroup;
use differential_dataflow::lattice::Lattice;
use futures_util::future::BoxFuture;
use futures_util::FutureExt;
use mz_persist::location::SeqNo;
use mz_persist_types::{Codec, Codec64};
use timely::progress::Timestamp;
use crate::internal::compact::CompactReq;
use crate::internal::gc::GcReq;
use crate::{Compactor, GarbageCollector, Machine};
/// Every handle to this shard may be occasionally asked to perform
/// routine maintenance after a successful compare_and_set operation.
///
/// For one-shot operations (like registering a reader) handles are
/// allowed to skip routine maintenance if necessary, as the same
/// maintenance operations will be recomputed by the next successful
/// compare_and_set of any handle.
///
/// Operations that run regularly once a handle is registered, such
/// as heartbeats, are expected to always perform maintenance.
#[must_use]
#[derive(Debug, Default, PartialEq)]
pub struct RoutineMaintenance {
pub(crate) garbage_collection: Option<GcReq>,
pub(crate) write_rollup: Option<SeqNo>,
}
impl RoutineMaintenance {
pub(crate) fn is_empty(&self) -> bool {
self == &RoutineMaintenance::default()
}
/// Initiates any routine maintenance necessary in background tasks
pub(crate) fn start_performing<K, V, T, D>(
self,
machine: &Machine<K, V, T, D>,
gc: &GarbageCollector<K, V, T, D>,
) where
K: Debug + Codec,
V: Debug + Codec,
T: Timestamp + Lattice + Codec64 + Sync,
D: Semigroup + Codec64 + Send + Sync,
{
let _ = self.perform_in_background(machine, gc);
}
/// Performs any routine maintenance necessary. Returns when all background
/// tasks have completed and the maintenance is done.
///
/// Generally, clients should prefer `start_performing` since it's typically not critical
/// that maintenance is performed by any particular deadline.
pub(crate) async fn perform<K, V, T, D>(
self,
machine: &Machine<K, V, T, D>,
gc: &GarbageCollector<K, V, T, D>,
) where
K: Debug + Codec,
V: Debug + Codec,
T: Timestamp + Lattice + Codec64 + Sync,
D: Semigroup + Codec64 + Send + Sync,
{
let mut more_maintenance = RoutineMaintenance::default();
for future in self.perform_in_background(machine, gc) {
more_maintenance.merge(future.await);
}
while !more_maintenance.is_empty() {
let maintenance = mem::take(&mut more_maintenance);
for future in maintenance.perform_in_background(machine, gc) {
more_maintenance.merge(future.await);
}
}
}
/// Initiates maintenance work in the background, either through spawned tasks
/// or by sending messages to existing tasks. The returned futures may be
/// awaited to know when the work is completed, but do not need to be polled
/// to drive the work to completion.
fn perform_in_background<K, V, T, D>(
self,
machine: &Machine<K, V, T, D>,
gc: &GarbageCollector<K, V, T, D>,
) -> Vec<BoxFuture<'static, RoutineMaintenance>>
where
K: Debug + Codec,
V: Debug + Codec,
T: Timestamp + Lattice + Codec64 + Sync,
D: Semigroup + Codec64 + Send + Sync,
{
let mut futures = vec![];
if let Some(gc_req) = self.garbage_collection {
if let Some(recv) = gc.gc_and_truncate_background(gc_req) {
// it's safe to ignore errors on the receiver. in the
// case of shutdown, the sender may have been dropped
futures.push(recv.map(Result::unwrap_or_default).boxed());
}
}
if let Some(rollup_seqno) = self.write_rollup {
let machine = machine.clone();
let isolated_runtime = Arc::clone(&machine.isolated_runtime);
futures.push(
isolated_runtime
.spawn_named(|| "persist::write_rollup", async move {
machine
.applier
.fetch_and_update_state(Some(rollup_seqno))
.await;
// We don't have to write at exactly rollup_seqno, just need
// something recent.
assert!(
machine.seqno() >= rollup_seqno,
"{} vs {}",
machine.seqno(),
rollup_seqno
);
machine.add_rollup_for_current_seqno().await
})
.map(Result::unwrap_or_default)
.boxed(),
);
}
futures
}
/// Merges another maintenance request into this one.
///
/// `other` is expected to come "later", and so its maintenance might
/// override `self`'s maintenance.
pub fn merge(&mut self, other: RoutineMaintenance) {
// Deconstruct other so we get a compile failure if new fields are
// added.
let RoutineMaintenance {
garbage_collection,
write_rollup,
} = other;
if let Some(garbage_collection) = garbage_collection {
self.garbage_collection = Some(garbage_collection);
}
if let Some(write_rollup) = write_rollup {
self.write_rollup = Some(write_rollup);
}
}
}
/// Writers may be asked to perform additional tasks beyond the
/// routine maintenance common to all handles. It is expected that
/// writers always perform maintenance.
#[must_use]
#[derive(Debug)]
pub struct WriterMaintenance<T> {
pub(crate) routine: RoutineMaintenance,
pub(crate) compaction: Vec<CompactReq<T>>,
}
impl<T> Default for WriterMaintenance<T> {
fn default() -> Self {
Self {
routine: RoutineMaintenance::default(),
compaction: Vec::default(),
}
}
}
impl<T> WriterMaintenance<T>
where
T: Timestamp + Lattice + Codec64 + Sync,
{
/// Initiates any writer maintenance necessary in background tasks
pub(crate) fn start_performing<K, V, D>(
self,
machine: &Machine<K, V, T, D>,
gc: &GarbageCollector<K, V, T, D>,
compactor: Option<&Compactor<K, V, T, D>>,
) where
K: Debug + Codec,
V: Debug + Codec,
D: Semigroup + Ord + Codec64 + Send + Sync,
{
let machine = machine.clone();
let gc = gc.clone();
let compactor = compactor.cloned();
mz_ore::task::spawn(|| "writer-maintenance", async move {
self.perform(&machine, &gc, compactor.as_ref()).await
});
}
/// Performs any writer maintenance necessary. Returns when all background
/// tasks have completed and the maintenance is done.
pub(crate) async fn perform<K, V, D>(
self,
machine: &Machine<K, V, T, D>,
gc: &GarbageCollector<K, V, T, D>,
compactor: Option<&Compactor<K, V, T, D>>,
) where
K: Debug + Codec,
V: Debug + Codec,
D: Semigroup + Ord + Codec64 + Send + Sync,
{
let Self {
routine,
compaction,
} = self;
let mut more_maintenance = RoutineMaintenance::default();
for future in routine.perform_in_background(machine, gc) {
more_maintenance.merge(future.await);
}
if let Some(compactor) = compactor {
for req in compaction {
if let Some(receiver) = compactor.compact_and_apply_background(req, machine) {
// it's safe to ignore errors on the receiver. in the
// case of shutdown, the sender may have been dropped
let _ = receiver.await;
}
}
}
while !more_maintenance.is_empty() {
let maintenance = mem::take(&mut more_maintenance);
for future in maintenance.perform_in_background(machine, gc) {
more_maintenance.merge(future.await);
}
}
}
}