Move the SQLite default constants (max connections, busy timeout) to
sqldb/config.go as the single source of truth and export them. Remove
the duplicate definitions from lncfg/db.go and reference the sqldb
constants instead.
Both the sqldb and kvdb SQLite layers were not using a sensible default
for MaxConnections. The sqldb store used defaultMaxConns (25) which is
meant for Postgres, and the kvdb path passed 0 (unlimited) to
sqlbase.Init when unconfigured.
Add a MaxConns() method on SqliteConfig that returns the configured
value or defaults to 2, appropriate for SQLite's single-writer model.
Use it in both sqldb/sqlite.go and lncfg/db.go so both layers share
the same default.
Replace hardcoded WithGlobalLock assignment with configurable
options wallet postgres backends. Also add the WithGlobalLock
option to the channeldb table for postgres backends.
Defaults:
- channeldb: false (allow concurrent access)
- wallet: true (maintain safe single-writer behavior)
Users can now override these defaults via:
- db.postgres.channeldb-with-global-lock
- db.postgres.walletdb-with-global-lock
This gives operators flexibility while maintaining safe defaults
until full native SQL migration is complete.
Moreover exclude db.postgres.walletdb-with-global-lock check
in the sample config file script. We cannot easily check the
correct default because we set it later in the LND startup
sequence so we exclude it.
This commit adds a BenchmarkFindOptimalSQLQueryConfig test in the
graph/db package which runs ForEachNode and ForEachChannel queries
against a local backend using various different values for the sql
QueryConfig struct. This is done to determine good default values to
use for the config options for sqlite vs postgres.
Now that the SQL query config values are configurable, we add some
validation to make sure that the user doesnt set a max batch size that
is larger than the limits for sqlite/postgres that have been determined
by the TestSQLSliceQueries test.
Here, we make the sql query option params (batch size and pagination
size) configurable. The defaults for SQLite vs Postgres are still the
same but will be changed in an upcoming commit.
In future releases we will use this flag to make potentially failing kv
to native SQL migrations optional. For this reason it is better to
rename it to avoid confusion.
This commit separates the execution of SQL and in-code migrations
from their construction. This change is necessary because,
currently, the SQL schema is migrated during the construction
phase in the lncfg package. However, migrations are typically
executed when individual stores are constructed within the
configuration builder.
This commit introduces support for custom, in-code migrations, allowing
a specific Go function to be executed at a designated database version
during sqlc migrations. If the current database version surpasses the
specified version, the migration will be skipped.
Find and replace all nolint instances refering to the `lll` linter and
replace with `ll` which is the name of our custom version of the `lll`
linter which can be used to ignore log lines during linting.
The next commit will do the configuration of the custom linter and
disable the default one.
This commit adds an optional separate "native" SQL backend that will
host all tables supporting native SQL. For now since we don't have many
such tables we'll keep it in one file for SQLite, but it may be split up
if desired.
In this commit, a new `--db.no-rev-log-amt-data` flag is added. The
config option is passed though to everywhere that it will be used. Note
that it is still a no-op in this commit. An upcoming commit will make
use of the flag.
Use kvdb package v1.4.1. This update also forced the protobuf version to
be bumped which required `make rpc` to be run to update the generated
files. This also required a bump in the github pinned dependencies
config for the grpc and protobuf libs.
In this commit we add an extra config for enabling logging to an
external file when using embedded etcd. This can be useful when running
integration tests to see more details about etcd related issues.
Since we're now storing the content of multiple previously distinct
database files in etcd, we want to properly namespace them as not to
provoke any key collisions. We append a sub namespace to the given
global namespace in order to still support multiple lnd nodes using the
same etcd database simultaneously.
Because the btcwallet code uses the legacy walletdb interface we must
assume it is not fully concurrency safe. Therefore we make sure only a
single writer can be active at any given time for the wallet DB backend
when using etcd.
To have all the database backend related code in one place, we finally
also move the initialization of the wallet DB loader option into the
GetBackends() method.
The final database that needs to be made remote compatible is the
watchtower server and client database.
They are handled a bit differently because both of them are not always
active, only when specifically turned on in the config.
Even though the sphinx router's persistent replay log is not crucial in
the operation of lnd as its state can be re-created by creating a new
brontide connection, we want to make lnd fully stateless and therefore
have the option of not storing any state on disk.
The macaroon root keys should also be stored to the remote database if a
replicated backend such as etcd is used.
This commit refactors the macaroons service and wallet unlocker to
accept a kvdb backend directly instead of creating the bolt instance
automatically.
This commit gets rid of the concept of a local and remote database when
etcd is used. Instead the same backend is now used for both the
(previously renamed from local and remote DB) graph and channel state
databases.
This will make path finding extremely slow on etcd and will require
further optimizations and possibly a write-through cache for the graph
DB. But this is a requirement for making lnd itself fully stateless.
In order to separate our databases more clearly, we refactor the height
hint cache DB to use a kvdb backend instead of the channel DB instance
directly.
As a preparation to not have a local and remote version of the database
around anymore, we rename the variables into what their actual function
is. In case of the RPC server we even directly use the channel graph
instead of the DB instance. This should allow us to extract the channel
graph into its own, separate database (perhaps with better access
characteristics) in the future.
ziggie