atuin/crates/atuin-client/src/kv.rs

266 lines
7.5 KiB
Rust
Raw Normal View History

use std::collections::BTreeMap;
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
use atuin_common::record::{DecryptedData, Host, HostId};
use eyre::{bail, ensure, eyre, Result};
use serde::Deserialize;
use crate::record::encryption::PASETO_V4;
use crate::record::store::Store;
const KV_VERSION: &str = "v0";
const KV_TAG: &str = "kv";
const KV_VAL_MAX_LEN: usize = 100 * 1024;
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct KvRecord {
2023-06-15 09:17:30 +02:00
pub namespace: String,
pub key: String,
pub value: String,
}
impl KvRecord {
pub fn serialize(&self) -> Result<DecryptedData> {
use rmp::encode;
let mut output = vec![];
// INFO: ensure this is updated when adding new fields
encode::write_array_len(&mut output, 3)?;
encode::write_str(&mut output, &self.namespace)?;
encode::write_str(&mut output, &self.key)?;
encode::write_str(&mut output, &self.value)?;
Ok(DecryptedData(output))
}
pub fn deserialize(data: &DecryptedData, version: &str) -> Result<Self> {
use rmp::decode;
fn error_report<E: std::fmt::Debug>(err: E) -> eyre::Report {
eyre!("{err:?}")
}
match version {
KV_VERSION => {
let mut bytes = decode::Bytes::new(&data.0);
let nfields = decode::read_array_len(&mut bytes).map_err(error_report)?;
ensure!(nfields == 3, "too many entries in v0 kv record");
let bytes = bytes.remaining_slice();
let (namespace, bytes) =
decode::read_str_from_slice(bytes).map_err(error_report)?;
let (key, bytes) = decode::read_str_from_slice(bytes).map_err(error_report)?;
let (value, bytes) = decode::read_str_from_slice(bytes).map_err(error_report)?;
if !bytes.is_empty() {
bail!("trailing bytes in encoded kvrecord. malformed")
}
Ok(KvRecord {
namespace: namespace.to_owned(),
key: key.to_owned(),
value: value.to_owned(),
})
}
_ => {
bail!("unknown version {version:?}")
}
}
}
}
#[derive(Debug, Clone, Deserialize)]
pub struct KvStore;
impl Default for KvStore {
fn default() -> Self {
Self::new()
}
}
impl KvStore {
// will want to init the actual kv store when that is done
pub fn new() -> KvStore {
KvStore {}
}
pub async fn set(
&self,
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
store: &(impl Store + Send + Sync),
encryption_key: &[u8; 32],
host_id: HostId,
2023-06-15 09:17:30 +02:00
namespace: &str,
key: &str,
value: &str,
) -> Result<()> {
if value.len() > KV_VAL_MAX_LEN {
return Err(eyre!(
"kv value too large: max len {} bytes",
KV_VAL_MAX_LEN
));
}
let record = KvRecord {
2023-06-15 09:17:30 +02:00
namespace: namespace.to_string(),
key: key.to_string(),
value: value.to_string(),
};
let bytes = record.serialize()?;
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
let idx = store
.last(host_id, KV_TAG)
.await?
.map_or(0, |entry| entry.idx + 1);
let record = atuin_common::record::Record::builder()
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
.host(Host::new(host_id))
.version(KV_VERSION.to_string())
.tag(KV_TAG.to_string())
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
.idx(idx)
.data(bytes)
.build();
store
.push(&record.encrypt::<PASETO_V4>(encryption_key))
.await?;
Ok(())
}
// TODO: setup an actual kv store, rebuild func, and do not pass the main store in here as
// well.
2023-06-15 09:17:30 +02:00
pub async fn get(
&self,
store: &impl Store,
encryption_key: &[u8; 32],
2023-06-15 09:17:30 +02:00
namespace: &str,
key: &str,
) -> Result<Option<KvRecord>> {
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
// TODO: don't rebuild every time...
let map = self.build_kv(store, encryption_key).await?;
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
let res = map.get(namespace);
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
if let Some(ns) = res {
let value = ns.get(key);
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
Ok(value.cloned())
} else {
Ok(None)
}
}
// Build a kv map out of the linked list kv store
// Map is Namespace -> Key -> Value
// TODO(ellie): "cache" this into a real kv structure, which we can
// use as a write-through cache to avoid constant rebuilds.
pub async fn build_kv(
&self,
store: &impl Store,
encryption_key: &[u8; 32],
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
) -> Result<BTreeMap<String, BTreeMap<String, KvRecord>>> {
let mut map = BTreeMap::new();
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
// TODO: maybe don't load the entire tag into memory to build the kv
// we can be smart about it and only load values since the last build
// or, iterate/paginate
let tagged = store.all_tagged(KV_TAG).await?;
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
// iterate through all tags and play each KV record at a time
// this is "last write wins"
// probably good enough for now, but revisit in future
for record in tagged {
let decrypted = match record.version.as_str() {
KV_VERSION => record.decrypt::<PASETO_V4>(encryption_key)?,
version => bail!("unknown version {version:?}"),
};
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
let kv = KvRecord::deserialize(&decrypted.data, KV_VERSION)?;
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
let ns = map
.entry(kv.namespace.clone())
.or_insert_with(BTreeMap::new);
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
ns.insert(kv.key.clone(), kv);
}
Ok(map)
}
}
#[cfg(test)]
mod tests {
2023-08-18 22:45:29 +02:00
use crypto_secretbox::{KeyInit, XSalsa20Poly1305};
use rand::rngs::OsRng;
use crate::record::sqlite_store::{test_sqlite_store_timeout, SqliteStore};
use super::{KvRecord, KvStore, KV_VERSION};
#[test]
fn encode_decode() {
let kv = KvRecord {
namespace: "foo".to_owned(),
key: "bar".to_owned(),
value: "baz".to_owned(),
};
let snapshot = [
0x93, 0xa3, b'f', b'o', b'o', 0xa3, b'b', b'a', b'r', 0xa3, b'b', b'a', b'z',
];
let encoded = kv.serialize().unwrap();
let decoded = KvRecord::deserialize(&encoded, KV_VERSION).unwrap();
assert_eq!(encoded.0, &snapshot);
assert_eq!(decoded, kv);
}
#[tokio::test]
async fn build_kv() {
let mut store = SqliteStore::new(":memory:", test_sqlite_store_timeout())
.await
.unwrap();
let kv = KvStore::new();
let key: [u8; 32] = XSalsa20Poly1305::generate_key(&mut OsRng).into();
let host_id = atuin_common::record::HostId(atuin_common::utils::uuid_v7());
kv.set(&mut store, &key, host_id, "test-kv", "foo", "bar")
.await
.unwrap();
kv.set(&mut store, &key, host_id, "test-kv", "1", "2")
.await
.unwrap();
let map = kv.build_kv(&store, &key).await.unwrap();
assert_eq!(
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
*map.get("test-kv")
.expect("map namespace not set")
.get("foo")
.expect("map key not set"),
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
KvRecord {
namespace: String::from("test-kv"),
key: String::from("foo"),
value: String::from("bar")
}
);
assert_eq!(
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
*map.get("test-kv")
.expect("map namespace not set")
.get("1")
.expect("map key not set"),
feat: rework record sync for improved reliability (#1478) * feat: rework record sync for improved reliability So, to tell a story 1. We introduced the record sync, intended to be the new algorithm to sync history. 2. On top of this, I added the KV store. This was intended as a simple test of the record sync, and to see if people wanted that sort of functionality 3. History remained syncing via the old means, as while it had issues it worked more-or-less OK. And we are aware of its flaws 4. If KV syncing worked ok, history would be moved across KV syncing ran ok for 6mo or so, so I started to move across history. For several weeks, I ran a local fork of Atuin + the server that synced via records instead. The record store maintained ordering via a linked list, which was a mistake. It performed well in testing, but was really difficult to debug and reason about. So when a few small sync issues occured, they took an extremely long time to debug. This PR is huge, which I regret. It involves replacing the "parent" relationship that records once had (pointing to the previous record) with a simple index (generally referred to as idx). This also means we had to change the recordindex, which referenced "tails". Tails were the last item in the chain. Now that we use an "array" vs linked list, that logic was also replaced. And is much simpler :D Same for the queries that act on this data. ---- This isn't final - we still need to add 1. Proper server/client error handling, which has been lacking for a while 2. The actual history implementation on top This exists in a branch, just without deletions. Won't be much to add that, I just don't want to make this any larger than it already is The _only_ caveat here is that we basically lose data synced via the old record store. This is the KV data from before. It hasn't been deleted or anything, just no longer hooked up. So it's totally possible to write a migration script. I just need to do that. * update .gitignore * use correct endpoint * fix for stores with length of 1 * use create/delete enum for history store * lint, remove unneeded host_id * remove prints * add command to import old history * add enable/disable switch for record sync * add record sync to auto sync * satisfy the almighty clippy * remove file that I did not mean to commit * feedback
2024-01-05 18:57:49 +01:00
KvRecord {
namespace: String::from("test-kv"),
key: String::from("1"),
value: String::from("2")
}
);
}
}