Remove epidemic propagation for fully replicated stuff: write directly to all nodes

This commit is contained in:
Alex Auvolat 2021-03-05 15:09:18 +01:00
parent d7e148d302
commit 3882d5ba36
7 changed files with 71 additions and 94 deletions

View file

@ -334,17 +334,28 @@ impl BlockManager {
} }
if need_nodes.len() > 0 { if need_nodes.len() > 0 {
trace!("Block {:?} needed by {} nodes, sending", hash, need_nodes.len()); trace!(
"Block {:?} needed by {} nodes, sending",
hash,
need_nodes.len()
);
let put_block_message = Arc::new(self.read_block(hash).await?); let put_block_message = Arc::new(self.read_block(hash).await?);
let put_resps = join_all(need_nodes.iter().map(|to| { let put_resps = join_all(need_nodes.iter().map(|to| {
self.rpc_client.call_arc(*to, put_block_message.clone(), BLOCK_RW_TIMEOUT) self.rpc_client
})).await; .call_arc(*to, put_block_message.clone(), BLOCK_RW_TIMEOUT)
}))
.await;
for resp in put_resps { for resp in put_resps {
resp?; resp?;
} }
} }
trace!("Deleting block {:?}, offload finished ({} / {})", hash, need_nodes.len(), who.len()); trace!(
"Deleting block {:?}, offload finished ({} / {})",
hash,
need_nodes.len(),
who.len()
);
fs::remove_file(path).await?; fs::remove_file(path).await?;
self.resync_queue.remove(&hash)?; self.resync_queue.remove(&hash)?;
@ -391,7 +402,7 @@ impl BlockManager {
.try_call_many( .try_call_many(
&who[..], &who[..],
Message::PutBlock(PutBlockMessage { hash, data }), Message::PutBlock(PutBlockMessage { hash, data }),
RequestStrategy::with_quorum(self.replication.write_quorum()) RequestStrategy::with_quorum(self.replication.write_quorum(&self.system))
.with_timeout(BLOCK_RW_TIMEOUT), .with_timeout(BLOCK_RW_TIMEOUT),
) )
.await?; .await?;
@ -420,12 +431,17 @@ impl BlockManager {
} }
// 2. Repair blocks actually on disk // 2. Repair blocks actually on disk
self.repair_aux_read_dir_rec(&self.data_dir, must_exit).await?; self.repair_aux_read_dir_rec(&self.data_dir, must_exit)
.await?;
Ok(()) Ok(())
} }
fn repair_aux_read_dir_rec<'a>(&'a self, path: &'a PathBuf, must_exit: &'a watch::Receiver<bool>) -> BoxFuture<'a, Result<(), Error>> { fn repair_aux_read_dir_rec<'a>(
&'a self,
path: &'a PathBuf,
must_exit: &'a watch::Receiver<bool>,
) -> BoxFuture<'a, Result<(), Error>> {
// Lists all blocks on disk and adds them to the resync queue. // Lists all blocks on disk and adds them to the resync queue.
// This allows us to find blocks we are storing but don't actually need, // This allows us to find blocks we are storing but don't actually need,
// so that we can offload them if necessary and then delete them locally. // so that we can offload them if necessary and then delete them locally.
@ -441,7 +457,8 @@ impl BlockManager {
let ent_type = data_dir_ent.file_type().await?; let ent_type = data_dir_ent.file_type().await?;
if name.len() == 2 && hex::decode(&name).is_ok() && ent_type.is_dir() { if name.len() == 2 && hex::decode(&name).is_ok() && ent_type.is_dir() {
self.repair_aux_read_dir_rec(&data_dir_ent.path(), must_exit).await?; self.repair_aux_read_dir_rec(&data_dir_ent.path(), must_exit)
.await?;
} else if name.len() == 64 { } else if name.len() == 64 {
let hash_bytes = match hex::decode(&name) { let hash_bytes = match hex::decode(&name) {
Ok(h) => h, Ok(h) => h,
@ -457,7 +474,8 @@ impl BlockManager {
} }
} }
Ok(()) Ok(())
}.boxed() }
.boxed()
} }
} }

View file

@ -65,10 +65,7 @@ impl Garage {
read_quorum: (config.meta_replication_factor + 1) / 2, read_quorum: (config.meta_replication_factor + 1) / 2,
}; };
let control_rep_param = TableFullReplication::new( let control_rep_param = TableFullReplication::new(config.control_write_max_faults);
config.meta_epidemic_fanout,
(config.meta_epidemic_fanout + 1) / 2,
);
info!("Initialize block manager..."); info!("Initialize block manager...");
let block_manager = BlockManager::new( let block_manager = BlockManager::new(

View file

@ -61,9 +61,8 @@ pub trait TableReplication: Send + Sync {
// Which nodes to send writes to // Which nodes to send writes to
fn write_nodes(&self, hash: &Hash, system: &System) -> Vec<UUID>; fn write_nodes(&self, hash: &Hash, system: &System) -> Vec<UUID>;
fn write_quorum(&self) -> usize; fn write_quorum(&self, system: &System) -> usize;
fn max_write_errors(&self) -> usize; fn max_write_errors(&self) -> usize;
fn epidemic_writes(&self) -> bool;
// Which are the nodes that do actually replicate the data // Which are the nodes that do actually replicate the data
fn replication_nodes(&self, hash: &Hash, ring: &Ring) -> Vec<UUID>; fn replication_nodes(&self, hash: &Hash, ring: &Ring) -> Vec<UUID>;
@ -119,7 +118,7 @@ where
.try_call_many( .try_call_many(
&who[..], &who[..],
rpc, rpc,
RequestStrategy::with_quorum(self.replication.write_quorum()) RequestStrategy::with_quorum(self.replication.write_quorum(&self.system))
.with_timeout(TABLE_RPC_TIMEOUT), .with_timeout(TABLE_RPC_TIMEOUT),
) )
.await?; .await?;
@ -382,7 +381,6 @@ where
pub async fn handle_update(self: &Arc<Self>, entries: &[Arc<ByteBuf>]) -> Result<(), Error> { pub async fn handle_update(self: &Arc<Self>, entries: &[Arc<ByteBuf>]) -> Result<(), Error> {
let syncer = self.syncer.load_full().unwrap(); let syncer = self.syncer.load_full().unwrap();
let mut epidemic_propagate = vec![];
for update_bytes in entries.iter() { for update_bytes in entries.iter() {
let update = self.decode_entry(update_bytes.as_slice())?; let update = self.decode_entry(update_bytes.as_slice())?;
@ -410,22 +408,11 @@ where
})?; })?;
if old_entry.as_ref() != Some(&new_entry) { if old_entry.as_ref() != Some(&new_entry) {
if self.replication.epidemic_writes() {
epidemic_propagate.push(new_entry.clone());
}
self.instance.updated(old_entry, Some(new_entry)); self.instance.updated(old_entry, Some(new_entry));
syncer.invalidate(&tree_key[..]); syncer.invalidate(&tree_key[..]);
} }
} }
if epidemic_propagate.len() > 0 {
let self2 = self.clone();
self.system
.background
.spawn_cancellable(async move { self2.insert_many(&epidemic_propagate[..]).await });
}
Ok(()) Ok(())
} }

View file

@ -1,4 +1,3 @@
use arc_swap::ArcSwapOption;
use std::sync::Arc; use std::sync::Arc;
use garage_rpc::membership::System; use garage_rpc::membership::System;
@ -9,10 +8,7 @@ use crate::*;
#[derive(Clone)] #[derive(Clone)]
pub struct TableFullReplication { pub struct TableFullReplication {
pub write_factor: usize, pub max_faults: usize,
pub write_quorum: usize,
neighbors: ArcSwapOption<Neighbors>,
} }
#[derive(Clone)] #[derive(Clone)]
@ -22,45 +18,8 @@ struct Neighbors {
} }
impl TableFullReplication { impl TableFullReplication {
pub fn new(write_factor: usize, write_quorum: usize) -> Self { pub fn new(max_faults: usize) -> Self {
TableFullReplication { TableFullReplication { max_faults }
write_factor,
write_quorum,
neighbors: ArcSwapOption::from(None),
}
}
fn get_neighbors(&self, system: &System) -> Vec<UUID> {
let neighbors = self.neighbors.load_full();
if let Some(n) = neighbors {
if Arc::ptr_eq(&n.ring, &system.ring.borrow()) {
return n.neighbors.clone();
}
}
// Recalculate neighbors
let ring = system.ring.borrow().clone();
let my_id = system.id;
let mut nodes = vec![];
for (node, _) in ring.config.members.iter() {
let node_ranking = fasthash(&[node.as_slice(), my_id.as_slice()].concat());
nodes.push((*node, node_ranking));
}
nodes.sort_by(|(_, rank1), (_, rank2)| rank1.cmp(rank2));
let mut neighbors = nodes
.drain(..)
.map(|(node, _)| node)
.filter(|node| *node != my_id)
.take(self.write_factor)
.collect::<Vec<_>>();
neighbors.push(my_id);
self.neighbors.swap(Some(Arc::new(Neighbors {
ring,
neighbors: neighbors.clone(),
})));
neighbors
} }
} }
@ -78,17 +37,14 @@ impl TableReplication for TableFullReplication {
1 1
} }
fn write_nodes(&self, _hash: &Hash, system: &System) -> Vec<UUID> { fn write_nodes(&self, hash: &Hash, system: &System) -> Vec<UUID> {
self.get_neighbors(system) self.replication_nodes(hash, system.ring.borrow().as_ref())
} }
fn write_quorum(&self) -> usize { fn write_quorum(&self, system: &System) -> usize {
self.write_quorum system.ring.borrow().config.members.len() - self.max_faults
} }
fn max_write_errors(&self) -> usize { fn max_write_errors(&self) -> usize {
self.write_factor - self.write_quorum self.max_faults
}
fn epidemic_writes(&self) -> bool {
true
} }
fn replication_nodes(&self, _hash: &Hash, ring: &Ring) -> Vec<UUID> { fn replication_nodes(&self, _hash: &Hash, ring: &Ring) -> Vec<UUID> {

View file

@ -31,15 +31,12 @@ impl TableReplication for TableShardedReplication {
let ring = system.ring.borrow().clone(); let ring = system.ring.borrow().clone();
ring.walk_ring(&hash, self.replication_factor) ring.walk_ring(&hash, self.replication_factor)
} }
fn write_quorum(&self) -> usize { fn write_quorum(&self, _system: &System) -> usize {
self.write_quorum self.write_quorum
} }
fn max_write_errors(&self) -> usize { fn max_write_errors(&self) -> usize {
self.replication_factor - self.write_quorum self.replication_factor - self.write_quorum
} }
fn epidemic_writes(&self) -> bool {
false
}
fn replication_nodes(&self, hash: &Hash, ring: &Ring) -> Vec<UUID> { fn replication_nodes(&self, hash: &Hash, ring: &Ring) -> Vec<UUID> {
ring.walk_ring(&hash, self.replication_factor) ring.walk_ring(&hash, self.replication_factor)

View file

@ -319,7 +319,13 @@ where
} }
counter += 1; counter += 1;
debug!("Offloading {} items from {:?}..{:?} ({})", items.len(), begin, end, counter); debug!(
"Offloading {} items from {:?}..{:?} ({})",
items.len(),
begin,
end,
counter
);
self.offload_items(&items, &nodes[..]).await?; self.offload_items(&items, &nodes[..]).await?;
} else { } else {
break; break;
@ -408,7 +414,11 @@ where
.iter() .iter()
.all(|x| *x == 0u8) .all(|x| *x == 0u8)
{ {
trace!("range_checksum {:?} returning {} items", range, children.len()); trace!(
"range_checksum {:?} returning {} items",
range,
children.len()
);
return Ok(RangeChecksum { return Ok(RangeChecksum {
bounds: range.clone(), bounds: range.clone(),
children, children,
@ -423,7 +433,11 @@ where
}; };
children.push((item_range, blake2sum(&value[..]))); children.push((item_range, blake2sum(&value[..])));
} }
trace!("range_checksum {:?} returning {} items", range, children.len()); trace!(
"range_checksum {:?} returning {} items",
range,
children.len()
);
Ok(RangeChecksum { Ok(RangeChecksum {
bounds: range.clone(), bounds: range.clone(),
children, children,
@ -449,7 +463,11 @@ where
} }
if sub_ck.found_limit.is_none() || sub_ck.hash.is_none() { if sub_ck.found_limit.is_none() || sub_ck.hash.is_none() {
trace!("range_checksum {:?} returning {} items", range, children.len()); trace!(
"range_checksum {:?} returning {} items",
range,
children.len()
);
return Ok(RangeChecksum { return Ok(RangeChecksum {
bounds: range.clone(), bounds: range.clone(),
children, children,
@ -464,7 +482,11 @@ where
.iter() .iter()
.all(|x| *x == 0u8) .all(|x| *x == 0u8)
{ {
trace!("range_checksum {:?} returning {} items", range, children.len()); trace!(
"range_checksum {:?} returning {} items",
range,
children.len()
);
return Ok(RangeChecksum { return Ok(RangeChecksum {
bounds: range.clone(), bounds: range.clone(),
children, children,

View file

@ -23,12 +23,12 @@ pub struct Config {
#[serde(default = "default_block_size")] #[serde(default = "default_block_size")]
pub block_size: usize, pub block_size: usize,
#[serde(default = "default_control_write_max_faults")]
pub control_write_max_faults: usize,
#[serde(default = "default_replication_factor")] #[serde(default = "default_replication_factor")]
pub meta_replication_factor: usize, pub meta_replication_factor: usize,
#[serde(default = "default_epidemic_fanout")]
pub meta_epidemic_fanout: usize,
#[serde(default = "default_replication_factor")] #[serde(default = "default_replication_factor")]
pub data_replication_factor: usize, pub data_replication_factor: usize,
@ -68,8 +68,8 @@ fn default_block_size() -> usize {
fn default_replication_factor() -> usize { fn default_replication_factor() -> usize {
3 3
} }
fn default_epidemic_fanout() -> usize { fn default_control_write_max_faults() -> usize {
3 1
} }
pub fn read_config(config_file: PathBuf) -> Result<Config, Error> { pub fn read_config(config_file: PathBuf) -> Result<Config, Error> {