Files
neon/libs/pageserver_api/src/shard.rs
Yuchen Liang a68edad913 refactor: move part of sharding API from pageserver_api to utils (#8254)
## Problem

LSN Leases introduced in #8084 is a new API that is made shard-aware
from day 1. To support ephemeral endpoint in #7994 without linking
Postgres C API against `compute_ctl`, part of the sharding needs to
reside in `utils`.

## Summary of changes

- Create a new `shard` module in utils crate.
- Move more interface related part of tenant sharding API to utils and
re-export them in pageserver_api.

Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-07-08 15:43:10 +01:00

594 lines
21 KiB
Rust

//! See docs/rfcs/031-sharding-static.md for an overview of sharding.
//!
//! This module contains a variety of types used to represent the concept of sharding
//! a Neon tenant across multiple physical shards. Since there are quite a few of these,
//! we provide an summary here.
//!
//! Types used to describe shards:
//! - [`ShardCount`] describes how many shards make up a tenant, plus the magic `unsharded` value
//! which identifies a tenant which is not shard-aware. This means its storage paths do not include
//! a shard suffix.
//! - [`ShardNumber`] is simply the zero-based index of a shard within a tenant.
//! - [`ShardIndex`] is the 2-tuple of `ShardCount` and `ShardNumber`, it's just like a `TenantShardId`
//! without the tenant ID. This is useful for things that are implicitly scoped to a particular
//! tenant, such as layer files.
//! - [`ShardIdentity`]` is the full description of a particular shard's parameters, in sufficient
//! detail to convert a [`Key`] to a [`ShardNumber`] when deciding where to write/read.
//! - The [`ShardSlug`] is a terse formatter for ShardCount and ShardNumber, written as
//! four hex digits. An unsharded tenant is `0000`.
//! - [`TenantShardId`] is the unique ID of a particular shard within a particular tenant
//!
//! Types used to describe the parameters for data distribution in a sharded tenant:
//! - [`ShardStripeSize`] controls how long contiguous runs of [`Key`]s (stripes) are when distributed across
//! multiple shards. Its value is given in 8kiB pages.
//! - [`ShardLayout`] describes the data distribution scheme, and at time of writing is
//! always zero: this is provided for future upgrades that might introduce different
//! data distribution schemes.
//!
//! Examples:
//! - A legacy unsharded tenant has one shard with ShardCount(0), ShardNumber(0), and its slug is 0000
//! - A single sharded tenant has one shard with ShardCount(1), ShardNumber(0), and its slug is 0001
//! - In a tenant with 4 shards, each shard has ShardCount(N), ShardNumber(i) where i in 0..N-1 (inclusive),
//! and their slugs are 0004, 0104, 0204, and 0304.
use crate::{key::Key, models::ShardParameters};
use postgres_ffi::relfile_utils::INIT_FORKNUM;
use serde::{Deserialize, Serialize};
#[doc(inline)]
pub use ::utils::shard::*;
/// The ShardIdentity contains enough information to map a [`Key`] to a [`ShardNumber`],
/// and to check whether that [`ShardNumber`] is the same as the current shard.
#[derive(Clone, Copy, Serialize, Deserialize, Eq, PartialEq, Debug)]
pub struct ShardIdentity {
pub number: ShardNumber,
pub count: ShardCount,
pub stripe_size: ShardStripeSize,
layout: ShardLayout,
}
/// Stripe size in number of pages
#[derive(Clone, Copy, Serialize, Deserialize, Eq, PartialEq, Debug)]
pub struct ShardStripeSize(pub u32);
impl Default for ShardStripeSize {
fn default() -> Self {
DEFAULT_STRIPE_SIZE
}
}
/// Layout version: for future upgrades where we might change how the key->shard mapping works
#[derive(Clone, Copy, Serialize, Deserialize, Eq, PartialEq, Debug)]
pub struct ShardLayout(u8);
const LAYOUT_V1: ShardLayout = ShardLayout(1);
/// ShardIdentity uses a magic layout value to indicate if it is unusable
const LAYOUT_BROKEN: ShardLayout = ShardLayout(255);
/// Default stripe size in pages: 256MiB divided by 8kiB page size.
const DEFAULT_STRIPE_SIZE: ShardStripeSize = ShardStripeSize(256 * 1024 / 8);
#[derive(thiserror::Error, Debug, PartialEq, Eq)]
pub enum ShardConfigError {
#[error("Invalid shard count")]
InvalidCount,
#[error("Invalid shard number")]
InvalidNumber,
#[error("Invalid stripe size")]
InvalidStripeSize,
}
impl ShardIdentity {
/// An identity with number=0 count=0 is a "none" identity, which represents legacy
/// tenants. Modern single-shard tenants should not use this: they should
/// have number=0 count=1.
pub const fn unsharded() -> Self {
Self {
number: ShardNumber(0),
count: ShardCount(0),
layout: LAYOUT_V1,
stripe_size: DEFAULT_STRIPE_SIZE,
}
}
/// A broken instance of this type is only used for `TenantState::Broken` tenants,
/// which are constructed in code paths that don't have access to proper configuration.
///
/// A ShardIdentity in this state may not be used for anything, and should not be persisted.
/// Enforcement is via assertions, to avoid making our interface fallible for this
/// edge case: it is the Tenant's responsibility to avoid trying to do any I/O when in a broken
/// state, and by extension to avoid trying to do any page->shard resolution.
pub fn broken(number: ShardNumber, count: ShardCount) -> Self {
Self {
number,
count,
layout: LAYOUT_BROKEN,
stripe_size: DEFAULT_STRIPE_SIZE,
}
}
/// The "unsharded" value is distinct from simply having a single shard: it represents
/// a tenant which is not shard-aware at all, and whose storage paths will not include
/// a shard suffix.
pub fn is_unsharded(&self) -> bool {
self.number == ShardNumber(0) && self.count == ShardCount(0)
}
/// Count must be nonzero, and number must be < count. To construct
/// the legacy case (count==0), use Self::unsharded instead.
pub fn new(
number: ShardNumber,
count: ShardCount,
stripe_size: ShardStripeSize,
) -> Result<Self, ShardConfigError> {
if count.0 == 0 {
Err(ShardConfigError::InvalidCount)
} else if number.0 > count.0 - 1 {
Err(ShardConfigError::InvalidNumber)
} else if stripe_size.0 == 0 {
Err(ShardConfigError::InvalidStripeSize)
} else {
Ok(Self {
number,
count,
layout: LAYOUT_V1,
stripe_size,
})
}
}
/// For use when creating ShardIdentity instances for new shards, where a creation request
/// specifies the ShardParameters that apply to all shards.
pub fn from_params(number: ShardNumber, params: &ShardParameters) -> Self {
Self {
number,
count: params.count,
layout: LAYOUT_V1,
stripe_size: params.stripe_size,
}
}
fn is_broken(&self) -> bool {
self.layout == LAYOUT_BROKEN
}
pub fn get_shard_number(&self, key: &Key) -> ShardNumber {
assert!(!self.is_broken());
key_to_shard_number(self.count, self.stripe_size, key)
}
/// Return true if the key should be ingested by this shard
///
/// Shards must ingest _at least_ keys which return true from this check.
pub fn is_key_local(&self, key: &Key) -> bool {
assert!(!self.is_broken());
if self.count < ShardCount(2) || (key_is_shard0(key) && self.number == ShardNumber(0)) {
true
} else {
key_to_shard_number(self.count, self.stripe_size, key) == self.number
}
}
/// Return true if the key should be discarded if found in this shard's
/// data store, e.g. during compaction after a split.
///
/// Shards _may_ drop keys which return false here, but are not obliged to.
pub fn is_key_disposable(&self, key: &Key) -> bool {
if key_is_shard0(key) {
// Q: Why can't we dispose of shard0 content if we're not shard 0?
// A1: because the WAL ingestion logic currently ingests some shard 0
// content on all shards, even though it's only read on shard 0. If we
// dropped it, then subsequent WAL ingest to these keys would encounter
// an error.
// A2: because key_is_shard0 also covers relation size keys, which are written
// on all shards even though they're only maintained accurately on shard 0.
false
} else {
!self.is_key_local(key)
}
}
/// Obtains the shard number and count combined into a `ShardIndex`.
pub fn shard_index(&self) -> ShardIndex {
ShardIndex {
shard_count: self.count,
shard_number: self.number,
}
}
pub fn shard_slug(&self) -> String {
if self.count > ShardCount(0) {
format!("-{:02x}{:02x}", self.number.0, self.count.0)
} else {
String::new()
}
}
/// Convenience for checking if this identity is the 0th shard in a tenant,
/// for special cases on shard 0 such as ingesting relation sizes.
pub fn is_shard_zero(&self) -> bool {
self.number == ShardNumber(0)
}
}
/// Whether this key is always held on shard 0 (e.g. shard 0 holds all SLRU keys
/// in order to be able to serve basebackup requests without peer communication).
fn key_is_shard0(key: &Key) -> bool {
// To decide what to shard out to shards >0, we apply a simple rule that only
// relation pages are distributed to shards other than shard zero. Everything else gets
// stored on shard 0. This guarantees that shard 0 can independently serve basebackup
// requests, and any request other than those for particular blocks in relations.
//
// The only exception to this rule is "initfork" data -- this relates to postgres's UNLOGGED table
// type. These are special relations, usually with only 0 or 1 blocks, and we store them on shard 0
// because they must be included in basebackups.
let is_initfork = key.field5 == INIT_FORKNUM;
!key.is_rel_block_key() || is_initfork
}
/// Provide the same result as the function in postgres `hashfn.h` with the same name
fn murmurhash32(mut h: u32) -> u32 {
h ^= h >> 16;
h = h.wrapping_mul(0x85ebca6b);
h ^= h >> 13;
h = h.wrapping_mul(0xc2b2ae35);
h ^= h >> 16;
h
}
/// Provide the same result as the function in postgres `hashfn.h` with the same name
fn hash_combine(mut a: u32, mut b: u32) -> u32 {
b = b.wrapping_add(0x9e3779b9);
b = b.wrapping_add(a << 6);
b = b.wrapping_add(a >> 2);
a ^= b;
a
}
/// Where a Key is to be distributed across shards, select the shard. This function
/// does not account for keys that should be broadcast across shards.
///
/// The hashing in this function must exactly match what we do in postgres smgr
/// code. The resulting distribution of pages is intended to preserve locality within
/// `stripe_size` ranges of contiguous block numbers in the same relation, while otherwise
/// distributing data pseudo-randomly.
///
/// The mapping of key to shard is not stable across changes to ShardCount: this is intentional
/// and will be handled at higher levels when shards are split.
fn key_to_shard_number(count: ShardCount, stripe_size: ShardStripeSize, key: &Key) -> ShardNumber {
// Fast path for un-sharded tenants or broadcast keys
if count < ShardCount(2) || key_is_shard0(key) {
return ShardNumber(0);
}
// relNode
let mut hash = murmurhash32(key.field4);
// blockNum/stripe size
hash = hash_combine(hash, murmurhash32(key.field6 / stripe_size.0));
ShardNumber((hash % count.0 as u32) as u8)
}
/// For debugging, while not exposing the internals.
#[derive(Debug)]
#[allow(unused)] // used by debug formatting by pagectl
struct KeyShardingInfo {
shard0: bool,
shard_number: ShardNumber,
}
pub fn describe(
key: &Key,
shard_count: ShardCount,
stripe_size: ShardStripeSize,
) -> impl std::fmt::Debug {
KeyShardingInfo {
shard0: key_is_shard0(key),
shard_number: key_to_shard_number(shard_count, stripe_size, key),
}
}
#[cfg(test)]
mod tests {
use std::str::FromStr;
use utils::{id::TenantId, Hex};
use super::*;
const EXAMPLE_TENANT_ID: &str = "1f359dd625e519a1a4e8d7509690f6fc";
#[test]
fn tenant_shard_id_string() -> Result<(), hex::FromHexError> {
let example = TenantShardId {
tenant_id: TenantId::from_str(EXAMPLE_TENANT_ID).unwrap(),
shard_count: ShardCount(10),
shard_number: ShardNumber(7),
};
let encoded = format!("{example}");
let expected = format!("{EXAMPLE_TENANT_ID}-070a");
assert_eq!(&encoded, &expected);
let decoded = TenantShardId::from_str(&encoded)?;
assert_eq!(example, decoded);
Ok(())
}
#[test]
fn tenant_shard_id_binary() -> Result<(), hex::FromHexError> {
let example = TenantShardId {
tenant_id: TenantId::from_str(EXAMPLE_TENANT_ID).unwrap(),
shard_count: ShardCount(10),
shard_number: ShardNumber(7),
};
let encoded = bincode::serialize(&example).unwrap();
let expected: [u8; 18] = [
0x1f, 0x35, 0x9d, 0xd6, 0x25, 0xe5, 0x19, 0xa1, 0xa4, 0xe8, 0xd7, 0x50, 0x96, 0x90,
0xf6, 0xfc, 0x07, 0x0a,
];
assert_eq!(Hex(&encoded), Hex(&expected));
let decoded = bincode::deserialize(&encoded).unwrap();
assert_eq!(example, decoded);
Ok(())
}
#[test]
fn tenant_shard_id_backward_compat() -> Result<(), hex::FromHexError> {
// Test that TenantShardId can decode a TenantId in human
// readable form
let example = TenantId::from_str(EXAMPLE_TENANT_ID).unwrap();
let encoded = format!("{example}");
assert_eq!(&encoded, EXAMPLE_TENANT_ID);
let decoded = TenantShardId::from_str(&encoded)?;
assert_eq!(example, decoded.tenant_id);
assert_eq!(decoded.shard_count, ShardCount(0));
assert_eq!(decoded.shard_number, ShardNumber(0));
Ok(())
}
#[test]
fn tenant_shard_id_forward_compat() -> Result<(), hex::FromHexError> {
// Test that a legacy TenantShardId encodes into a form that
// can be decoded as TenantId
let example_tenant_id = TenantId::from_str(EXAMPLE_TENANT_ID).unwrap();
let example = TenantShardId::unsharded(example_tenant_id);
let encoded = format!("{example}");
assert_eq!(&encoded, EXAMPLE_TENANT_ID);
let decoded = TenantId::from_str(&encoded)?;
assert_eq!(example_tenant_id, decoded);
Ok(())
}
#[test]
fn tenant_shard_id_legacy_binary() -> Result<(), hex::FromHexError> {
// Unlike in human readable encoding, binary encoding does not
// do any special handling of legacy unsharded TenantIds: this test
// is equivalent to the main test for binary encoding, just verifying
// that the same behavior applies when we have used `unsharded()` to
// construct a TenantShardId.
let example = TenantShardId::unsharded(TenantId::from_str(EXAMPLE_TENANT_ID).unwrap());
let encoded = bincode::serialize(&example).unwrap();
let expected: [u8; 18] = [
0x1f, 0x35, 0x9d, 0xd6, 0x25, 0xe5, 0x19, 0xa1, 0xa4, 0xe8, 0xd7, 0x50, 0x96, 0x90,
0xf6, 0xfc, 0x00, 0x00,
];
assert_eq!(Hex(&encoded), Hex(&expected));
let decoded = bincode::deserialize::<TenantShardId>(&encoded).unwrap();
assert_eq!(example, decoded);
Ok(())
}
#[test]
fn shard_identity_validation() -> Result<(), ShardConfigError> {
// Happy cases
ShardIdentity::new(ShardNumber(0), ShardCount(1), DEFAULT_STRIPE_SIZE)?;
ShardIdentity::new(ShardNumber(0), ShardCount(1), ShardStripeSize(1))?;
ShardIdentity::new(ShardNumber(254), ShardCount(255), ShardStripeSize(1))?;
assert_eq!(
ShardIdentity::new(ShardNumber(0), ShardCount(0), DEFAULT_STRIPE_SIZE),
Err(ShardConfigError::InvalidCount)
);
assert_eq!(
ShardIdentity::new(ShardNumber(10), ShardCount(10), DEFAULT_STRIPE_SIZE),
Err(ShardConfigError::InvalidNumber)
);
assert_eq!(
ShardIdentity::new(ShardNumber(11), ShardCount(10), DEFAULT_STRIPE_SIZE),
Err(ShardConfigError::InvalidNumber)
);
assert_eq!(
ShardIdentity::new(ShardNumber(255), ShardCount(255), DEFAULT_STRIPE_SIZE),
Err(ShardConfigError::InvalidNumber)
);
assert_eq!(
ShardIdentity::new(ShardNumber(0), ShardCount(1), ShardStripeSize(0)),
Err(ShardConfigError::InvalidStripeSize)
);
Ok(())
}
#[test]
fn shard_index_human_encoding() -> Result<(), hex::FromHexError> {
let example = ShardIndex {
shard_number: ShardNumber(13),
shard_count: ShardCount(17),
};
let expected: String = "0d11".to_string();
let encoded = format!("{example}");
assert_eq!(&encoded, &expected);
let decoded = ShardIndex::from_str(&encoded)?;
assert_eq!(example, decoded);
Ok(())
}
#[test]
fn shard_index_binary_encoding() -> Result<(), hex::FromHexError> {
let example = ShardIndex {
shard_number: ShardNumber(13),
shard_count: ShardCount(17),
};
let expected: [u8; 2] = [0x0d, 0x11];
let encoded = bincode::serialize(&example).unwrap();
assert_eq!(Hex(&encoded), Hex(&expected));
let decoded = bincode::deserialize(&encoded).unwrap();
assert_eq!(example, decoded);
Ok(())
}
// These are only smoke tests to spot check that our implementation doesn't
// deviate from a few examples values: not aiming to validate the overall
// hashing algorithm.
#[test]
fn murmur_hash() {
assert_eq!(murmurhash32(0), 0);
assert_eq!(hash_combine(0xb1ff3b40, 0), 0xfb7923c9);
}
#[test]
fn shard_mapping() {
let key = Key {
field1: 0x00,
field2: 0x67f,
field3: 0x5,
field4: 0x400c,
field5: 0x00,
field6: 0x7d06,
};
let shard = key_to_shard_number(ShardCount(10), DEFAULT_STRIPE_SIZE, &key);
assert_eq!(shard, ShardNumber(8));
}
#[test]
fn shard_id_split() {
let tenant_id = TenantId::generate();
let parent = TenantShardId::unsharded(tenant_id);
// Unsharded into 2
assert_eq!(
parent.split(ShardCount(2)),
vec![
TenantShardId {
tenant_id,
shard_count: ShardCount(2),
shard_number: ShardNumber(0)
},
TenantShardId {
tenant_id,
shard_count: ShardCount(2),
shard_number: ShardNumber(1)
}
]
);
// Unsharded into 4
assert_eq!(
parent.split(ShardCount(4)),
vec![
TenantShardId {
tenant_id,
shard_count: ShardCount(4),
shard_number: ShardNumber(0)
},
TenantShardId {
tenant_id,
shard_count: ShardCount(4),
shard_number: ShardNumber(1)
},
TenantShardId {
tenant_id,
shard_count: ShardCount(4),
shard_number: ShardNumber(2)
},
TenantShardId {
tenant_id,
shard_count: ShardCount(4),
shard_number: ShardNumber(3)
}
]
);
// count=1 into 2 (check this works the same as unsharded.)
let parent = TenantShardId {
tenant_id,
shard_count: ShardCount(1),
shard_number: ShardNumber(0),
};
assert_eq!(
parent.split(ShardCount(2)),
vec![
TenantShardId {
tenant_id,
shard_count: ShardCount(2),
shard_number: ShardNumber(0)
},
TenantShardId {
tenant_id,
shard_count: ShardCount(2),
shard_number: ShardNumber(1)
}
]
);
// count=2 into count=8
let parent = TenantShardId {
tenant_id,
shard_count: ShardCount(2),
shard_number: ShardNumber(1),
};
assert_eq!(
parent.split(ShardCount(8)),
vec![
TenantShardId {
tenant_id,
shard_count: ShardCount(8),
shard_number: ShardNumber(1)
},
TenantShardId {
tenant_id,
shard_count: ShardCount(8),
shard_number: ShardNumber(3)
},
TenantShardId {
tenant_id,
shard_count: ShardCount(8),
shard_number: ShardNumber(5)
},
TenantShardId {
tenant_id,
shard_count: ShardCount(8),
shard_number: ShardNumber(7)
},
]
);
}
}