Files
neon/pageserver/src/basebackup.rs
Heikki Linnakangas 4f7b22a8a8 Refactor ObjectTags, intruducing a new concept called "relish"
This clarifies - I hope - the abstractions between Repository and
ObjectRepository. The ObjectTag struct was a mix of objects that could
be accessed directly through the public Timeline interface, and also
objects that were created and used internally by the ObjectRepository
implementation and not supposed to be accessed directly by the
callers.  With the RelishTag separaate from ObjectTag, the distinction
is more clear: RelishTag is used in the public interface, and
ObjectTag is used internally between object_repository.rs and
object_store.rs, and it contains the internal metadata object types.

One awkward thing with the ObjectTag struct was that the Repository
implementation had to distinguish between ObjectTags for relations,
and track the size of the relation, while others were used to store
"blobs".  With the RelishTags, some relishes are considered
"non-blocky", and the Repository implementation is expected to track
their sizes, while others are stored as blobs. I'm not 100% happy with
how RelishTag captures that either: it just knows that some relish
kinds are blocky and some non-blocky, and there's an is_block()
function to check that.  But this does enable size-tracking for SLRUs,
allowing us to treat them more like relations.

This changes the way SLRUs are stored in the repository. Each SLRU
segment, e.g. "pg_clog/0000", "pg_clog/0001", are now handled as a
separate relish.  This removes the need for the SLRU-specific
put_slru_truncate() function in the Timeline trait. SLRU truncation is
now handled by caling put_unlink() on the segment. This is more in
line with how PostgreSQL stores SLRUs and handles their trunction.

The SLRUs are "blocky", so they are accessed one 8k page at a time,
and repository tracks their size. I considered an alternative design
where we would treat each SLRU segment as non-blocky, and just store
the whole file as one blob. Each SLRU segment is up to 256 kB in size,
which isn't that large, so that might've worked fine, too. One reason
I didn't do that is that it seems better to have the WAL redo
routines be as close as possible to the PostgreSQL routines. It
doesn't matter much in the repository, though; we have to track the
size for relations anyway, so there's not much difference in whether
we also do it for SLRUs.
2021-08-01 13:16:16 +03:00

291 lines
10 KiB
Rust

//!
//! Generate a tarball with files needed to bootstrap ComputeNode.
//!
//! TODO: this module has nothing to do with PostgreSQL pg_basebackup.
//! It could use a better name.
//!
//! Stateless Postgres compute node is launched by sending a tarball
//! which contains non-relational data (multixacts, clog, filenodemaps, twophase files),
//! generated pg_control and dummy segment of WAL.
//! This module is responsible for creation of such tarball
//! from data stored in object storage.
//!
use bytes::{BufMut, BytesMut};
use log::*;
use std::io;
use std::io::Write;
use std::sync::Arc;
use std::time::SystemTime;
use tar::{Builder, EntryType, Header};
use crate::relish::*;
use crate::repository::Timeline;
use postgres_ffi::xlog_utils::*;
use postgres_ffi::*;
use zenith_utils::lsn::Lsn;
/// This is short-living object only for the time of tarball creation,
/// created mostly to avoid passing a lot of parameters between various functions
/// used for constructing tarball.
pub struct Basebackup<'a> {
ar: Builder<&'a mut dyn Write>,
timeline: &'a Arc<dyn Timeline>,
lsn: Lsn,
prev_record_lsn: Lsn,
}
impl<'a> Basebackup<'a> {
pub fn new(
write: &'a mut dyn Write,
timeline: &'a Arc<dyn Timeline>,
lsn: Lsn,
prev_record_lsn: Lsn,
) -> Basebackup<'a> {
Basebackup {
ar: Builder::new(write),
timeline,
lsn,
prev_record_lsn,
}
}
pub fn send_tarball(&mut self) -> anyhow::Result<()> {
// Create pgdata subdirs structure
for dir in pg_constants::PGDATA_SUBDIRS.iter() {
info!("send subdir {:?}", *dir);
let header = new_tar_header_dir(*dir)?;
self.ar.append(&header, &mut io::empty())?;
}
// Send empty config files.
for filepath in pg_constants::PGDATA_SPECIAL_FILES.iter() {
if *filepath == "pg_hba.conf" {
let data = pg_constants::PG_HBA.as_bytes();
let header = new_tar_header(&filepath, data.len() as u64)?;
self.ar.append(&header, &data[..])?;
} else {
let header = new_tar_header(&filepath, 0)?;
self.ar.append(&header, &mut io::empty())?;
}
}
// Gather non-relational files from object storage pages.
// Iteration is sorted order: all objects of the same type are grouped and traversed
// in key ascending order. For example all pg_xact records precede pg_multixact records and are sorted by block number.
// It allows to easily construct SLRU segments.
for obj in self.timeline.list_nonrels(self.lsn)? {
match obj {
RelishTag::Slru { slru, segno } => {
self.add_slru_segment(slru, segno)?;
}
RelishTag::FileNodeMap { spcnode, dbnode } => {
self.add_relmap_file(spcnode, dbnode)?;
}
RelishTag::TwoPhase { xid } => {
self.add_twophase_file(xid)?;
}
_ => {}
}
}
// Generate pg_control and bootstrap WAL segment.
self.add_pgcontrol_file()?;
self.ar.finish()?;
debug!("all tarred up!");
Ok(())
}
//
// Generate SLRU segment files from repository. Path identifies SLRU kind (pg_xact, pg_multixact/members, ...).
//
fn add_slru_segment(&mut self, slru: SlruKind, segno: u32) -> anyhow::Result<()> {
let nblocks = self
.timeline
.get_rel_size(RelishTag::Slru { slru, segno }, self.lsn)?;
let mut slru_buf: Vec<u8> =
Vec::with_capacity(nblocks as usize * pg_constants::BLCKSZ as usize);
for blknum in 0..nblocks {
let img = self.timeline.get_page_at_lsn_nowait(
RelishTag::Slru { slru, segno },
blknum,
self.lsn,
)?;
assert!(img.len() == pg_constants::BLCKSZ as usize);
slru_buf.extend_from_slice(&img);
}
let dir = match slru {
SlruKind::Clog => "pg_xact",
SlruKind::MultiXactMembers => "pg_multixact/members",
SlruKind::MultiXactOffsets => "pg_multixact/offsets",
};
let segname = format!("{}/{:>04X}", dir, segno);
let header = new_tar_header(&segname, slru_buf.len() as u64)?;
self.ar.append(&header, slru_buf.as_slice())?;
Ok(())
}
//
// Extract pg_filenode.map files from repository
// Along with them also send PG_VERSION for each database.
//
fn add_relmap_file(&mut self, spcnode: u32, dbnode: u32) -> anyhow::Result<()> {
let img = self.timeline.get_page_at_lsn_nowait(
RelishTag::FileNodeMap { spcnode, dbnode },
0,
self.lsn,
)?;
let path = if spcnode == pg_constants::GLOBALTABLESPACE_OID {
let dst_path = "PG_VERSION";
let version_bytes = pg_constants::PG_MAJORVERSION.as_bytes();
let header = new_tar_header(&dst_path, version_bytes.len() as u64)?;
self.ar.append(&header, &version_bytes[..])?;
let dst_path = format!("global/PG_VERSION");
let header = new_tar_header(&dst_path, version_bytes.len() as u64)?;
self.ar.append(&header, &version_bytes[..])?;
String::from("global/pg_filenode.map") // filenode map for global tablespace
} else {
// User defined tablespaces are not supported
assert!(spcnode == pg_constants::DEFAULTTABLESPACE_OID);
// Append dir path for each database
let path = format!("base/{}", dbnode);
let header = new_tar_header_dir(&path)?;
self.ar.append(&header, &mut io::empty())?;
let dst_path = format!("base/{}/PG_VERSION", dbnode);
let version_bytes = pg_constants::PG_MAJORVERSION.as_bytes();
let header = new_tar_header(&dst_path, version_bytes.len() as u64)?;
self.ar.append(&header, &version_bytes[..])?;
format!("base/{}/pg_filenode.map", dbnode)
};
assert!(img.len() == 512);
let header = new_tar_header(&path, img.len() as u64)?;
self.ar.append(&header, &img[..])?;
Ok(())
}
//
// Extract twophase state files
//
fn add_twophase_file(&mut self, xid: TransactionId) -> anyhow::Result<()> {
// Include in tarball two-phase files only of in-progress transactions
if self.timeline.get_tx_status(xid, self.lsn)?
== pg_constants::TRANSACTION_STATUS_IN_PROGRESS
{
let img =
self.timeline
.get_page_at_lsn_nowait(RelishTag::TwoPhase { xid }, 0, self.lsn)?;
let mut buf = BytesMut::new();
buf.extend_from_slice(&img[..]);
let crc = crc32c::crc32c(&img[..]);
buf.put_u32_le(crc);
let path = format!("pg_twophase/{:>08X}", xid);
let header = new_tar_header(&path, buf.len() as u64)?;
self.ar.append(&header, &buf[..])?;
}
Ok(())
}
//
// Add generated pg_control file and bootstrap WAL segment.
// Also send zenith.signal file with extra bootstrap data.
//
fn add_pgcontrol_file(&mut self) -> anyhow::Result<()> {
let checkpoint_bytes =
self.timeline
.get_page_at_lsn_nowait(RelishTag::Checkpoint, 0, self.lsn)?;
let pg_control_bytes =
self.timeline
.get_page_at_lsn_nowait(RelishTag::ControlFile, 0, self.lsn)?;
let mut pg_control = ControlFileData::decode(&pg_control_bytes)?;
let mut checkpoint = CheckPoint::decode(&checkpoint_bytes)?;
// Generate new pg_control and WAL needed for bootstrap
let checkpoint_segno = self.lsn.segment_number(pg_constants::WAL_SEGMENT_SIZE);
let checkpoint_lsn = XLogSegNoOffsetToRecPtr(
checkpoint_segno,
XLOG_SIZE_OF_XLOG_LONG_PHD as u32,
pg_constants::WAL_SEGMENT_SIZE,
);
checkpoint.redo = self.lsn.0 + self.lsn.calc_padding(8u32);
//reset some fields we don't want to preserve
//TODO Check this.
//We may need to determine the value from twophase data.
checkpoint.oldestActiveXid = 0;
//save new values in pg_control
pg_control.checkPoint = checkpoint_lsn;
pg_control.checkPointCopy = checkpoint;
pg_control.state = pg_constants::DB_SHUTDOWNED;
// add zenith.signal file
self.ar.append(
&new_tar_header("zenith.signal", 8)?,
&self.prev_record_lsn.0.to_le_bytes()[..],
)?;
//send pg_control
let pg_control_bytes = pg_control.encode();
let header = new_tar_header("global/pg_control", pg_control_bytes.len() as u64)?;
self.ar.append(&header, &pg_control_bytes[..])?;
//send wal segment
let wal_file_name = XLogFileName(
1, // FIXME: always use Postgres timeline 1
checkpoint_segno,
pg_constants::WAL_SEGMENT_SIZE,
);
let wal_file_path = format!("pg_wal/{}", wal_file_name);
let header = new_tar_header(&wal_file_path, pg_constants::WAL_SEGMENT_SIZE as u64)?;
let wal_seg = generate_wal_segment(&pg_control);
assert!(wal_seg.len() == pg_constants::WAL_SEGMENT_SIZE);
self.ar.append(&header, &wal_seg[..])?;
Ok(())
}
}
//
// Create new tarball entry header
//
fn new_tar_header(path: &str, size: u64) -> anyhow::Result<Header> {
let mut header = Header::new_gnu();
header.set_size(size);
header.set_path(path)?;
header.set_mode(0b110000000); // -rw-------
header.set_mtime(
// use currenttime as last modified time
SystemTime::now()
.duration_since(SystemTime::UNIX_EPOCH)
.unwrap()
.as_secs(),
);
header.set_cksum();
Ok(header)
}
fn new_tar_header_dir(path: &str) -> anyhow::Result<Header> {
let mut header = Header::new_gnu();
header.set_size(0);
header.set_path(path)?;
header.set_mode(0o755); // -rw-------
header.set_entry_type(EntryType::dir());
header.set_mtime(
// use currenttime as last modified time
SystemTime::now()
.duration_since(SystemTime::UNIX_EPOCH)
.unwrap()
.as_secs(),
);
header.set_cksum();
Ok(header)
}