|
| 1 | +# Data migration |
| 2 | + |
| 3 | +This document describes the process of migrating `LND`'s database state from one |
| 4 | +type of database backend (for example the `bbolt` based database files `*.db` |
| 5 | +such as the `channel.db` or `wallet.db` files) to another (for example the new |
| 6 | +`postgres` or `sqlite` databases introduced in `lnd v0.14.0-beta`). |
| 7 | + |
| 8 | +**Note:** Currently only migrations from `bolt` to either `postgres` or |
| 9 | +`sqlite` are supported. KV based databases will be phased out in the future. We |
| 10 | + are planning to add additonal support to migrate all `etcd` databases to `sql` |
| 11 | + databases. Moreover it is currently also not supported to move the database |
| 12 | + from `sqlite` to `postgres` or vice versa, because this migration only takes |
| 13 | + care of the key value data and there already exist the possiblity to have the |
| 14 | + invoices in native sql which this tool does not support. |
| 15 | + |
| 16 | + |
| 17 | +## Prepare the destination database |
| 18 | + |
| 19 | +To be able to execute the migration successfully you have to update to [LND |
| 20 | +v0.19.0](https://github.com/lightningnetwork/lnd/releases/tag/untagged-6824d65b23dc77e94e22) |
| 21 | +which makes sure that the intial pre-migration state of the kv-db is up to date. |
| 22 | + |
| 23 | +### Using postgres as the destination remote database |
| 24 | + |
| 25 | +Prepare a user and database as described in the [Postgres]( |
| 26 | + https://github.com/lightningnetwork/lnd/blob/master/docs/postgres.md) |
| 27 | +documentation. You'll need the Data Source Name (DSN) for both the data |
| 28 | +migration and then the `lnd` configuration, so keep that string somewhere |
| 29 | +(should be something with the format of `postgres://xx:yy@localhost:5432/zz`). |
| 30 | + |
| 31 | +No additional steps are required to prepare the Postgres database for the data |
| 32 | +migration. The migration tool will create the database schema automatically, so |
| 33 | +no DDL scripts need to be run in advance. But to speed up the migration process |
| 34 | +you should take a look at the following [postgres server tuning guide](https://gist.github.com/djkazic/526fa3e032aea9578997f88b45b91fb9) |
| 35 | + |
| 36 | +### Using sqlite as the destination remote database |
| 37 | + |
| 38 | +No particular preparation is needed for `sqlite` compared to the `postgres` |
| 39 | +case. Similar to the `bolt` case there will be separated db files created for |
| 40 | +each individual `bolt` database. |
| 41 | + |
| 42 | + |
| 43 | + |
| 44 | +## Prepare the source database |
| 45 | + |
| 46 | +Assuming we want to migrate the database state from the pre-0.19.0 individual |
| 47 | +`bbolt` based `*.db` files to a remote database, we first need to make sure the |
| 48 | +source files are in the correct state. |
| 49 | + |
| 50 | +The following steps should be performed *before* running the data migration ( |
| 51 | +some of them are marked as optional and can be neglected for small databases |
| 52 | +e.g. 200MB |
| 53 | +): |
| 54 | +1. Stop `lnd` |
| 55 | +2. Upgrade the `lnd` binary to the latest version (e.g. `v0.19.0-beta` or later) |
| 56 | +3. (optional) Make sure to add config options like |
| 57 | + `gc-canceled-invoices-on-startup=true` and `db.bolt.auto-compact=true` to |
| 58 | + your `lnd.conf` to optimize the source database size by removing canceled |
| 59 | + invoices and compacting it on startup. |
| 60 | +4. (optional) Also make sure to migrate the revocation log for all channels |
| 61 | + active prior to `[email protected]` by activating the config setting `--db.prune-revocation`. |
| 62 | + This version introduced an optimized revocation log storage system that |
| 63 | + reduces the storage footprint. All channels will be automatically migrated |
| 64 | + to this new format when the setting is enabled. |
| 65 | +5. Start `lnd` normally, using the flags mentioned above but not yet changing |
| 66 | + any database backend related configuration options. Check the log that the |
| 67 | + database schema was migrated successfully, for example: `Checking for |
| 68 | + schema update: latest_version=XX, db_version=XX`. This relates to the |
| 69 | + kv-schema NOT any SQL schema, this makes sure all inital migration of the |
| 70 | + old database were performed. |
| 71 | +6. (optional) Remove any data from the source database that you can. The fewer |
| 72 | + entries are in the source database, the quicker the migration will complete. |
| 73 | + For example failed payments (or their failed HTLC attempts) can be removed |
| 74 | + with `lncli deletepayments`. |
| 75 | +7. Stop `lnd` again and make sure it isn't started again by accident during the |
| 76 | + data migration (e.g. disable any `systemd` or other scripts that start/stop |
| 77 | + `lnd`). |
| 78 | + |
| 79 | +## Run the migration |
| 80 | + |
| 81 | +Depending on the destination database type, run the migration with a command |
| 82 | +similar to one of the following examples: |
| 83 | + |
| 84 | +**Example: Migrate from `bbolt` to `sqlite`:** |
| 85 | + |
| 86 | +```shell |
| 87 | +lndinit --debuglevel info migrate-db \ |
| 88 | +--source.bolt.data-dir /home/myuser/.lnd/data \ |
| 89 | +--dest.backend sqlite \ |
| 90 | +--dest.sqlite.data-dir /home/myuser/.lnd/data --network mainnet |
| 91 | +``` |
| 92 | +If you were running a watchtower server, and it had a different directory set |
| 93 | +compared to the default `LND` directory make sure you also add the tower dir |
| 94 | +setting. It has to be the directory to the `watchtower` dir, excluding the |
| 95 | +`watchtower` name e.g.: |
| 96 | +`--source.bolt.tower-dir /home/myuser/towerdir`. |
| 97 | + |
| 98 | + |
| 99 | +**Example: Migrate from `bbolt` to `postgres`:** |
| 100 | + |
| 101 | +```shell |
| 102 | +lndinit --debuglevel info migrate-db \ |
| 103 | + --source.bolt.data-dir /home/myuser/.lnd/data \ |
| 104 | + --dest.backend postgres \ |
| 105 | + --dest.postgres.dsn=postgres://postgres:postgres@localhost:5432/postgres |
| 106 | +``` |
| 107 | + |
| 108 | +Also set the watchtower directory in case you used a different path, see above. |
| 109 | + |
| 110 | +This migration tool depends on the directory structure of |
| 111 | +the LND software. This means make sure you link the correct folder because the |
| 112 | +`bolt` database has several database files in serveral subfolders. This |
| 113 | +migration tool will make sure that all the required databases are migrated. It |
| 114 | +will not require a `wtclient.db` or a `watchtower.db`. |
| 115 | + |
| 116 | +The migration is resumable and happens in chunks of default 20MB to make it |
| 117 | +compatible with most of the systems including low power devices like |
| 118 | +raspberry pis. However if you have better setup with way more RAM feel free to |
| 119 | +increase the `--chunk-size` to something like |
| 120 | +200MB which should speed up the migration significantly. You can also change |
| 121 | +the chunk size during the migration as well. |
| 122 | +If you want to start the migration from the beginning (can only be done if the |
| 123 | +migration did still not succeed) use the flag `--force-new-migration` which can |
| 124 | +be used in combination with a new `chunksize` limit. |
| 125 | + |
| 126 | + |
| 127 | +## After the migration was successful |
| 128 | + |
| 129 | +Make sure the whole migration process succeeds before starting the new node with |
| 130 | +the new underlying database. As mentioned above there are several database files |
| 131 | +at play here so all of them have to succeed to garantee a successful migration. |
| 132 | + |
| 133 | +If the migration succeeded successfully and you see the following log entry of |
| 134 | +the migration tool you are good to go to start-up your lnd node with the new |
| 135 | +database. |
| 136 | + |
| 137 | +```shell |
| 138 | +[INF]: LNDINIT !!!Migration of all mandatory db parts completed successfully!!! |
| 139 | +``` |
| 140 | + |
| 141 | +The mandatory dbs are: |
| 142 | +* `channel.db` |
| 143 | +* `macaroons.db` |
| 144 | +* `sphinxreplay.db` |
| 145 | +* `wallet.db` |
| 146 | + |
| 147 | +The optional dbs are: |
| 148 | + |
| 149 | +* `wtclient.db` |
| 150 | +* `watchtower.db` |
| 151 | +* `neutrino.db` |
| 152 | + |
| 153 | +### LND config setting for `sqlite` backend |
| 154 | + |
| 155 | +```shell |
| 156 | +[db] |
| 157 | +db.backend=sqlite |
| 158 | +``` |
| 159 | + |
| 160 | +There are several other sqlite setttings you can tweak to make it fit your |
| 161 | +needs, take a look at the [lnd-sample-config](https://github.com/lightningnetwork/lnd/blob/b6d8ecc7479f7517368814c398b0fbe0e8c52fed/sample-lnd.conf). |
| 162 | + |
| 163 | +### LND config setting for `postgres` backend |
| 164 | + |
| 165 | +```shell |
| 166 | +[db] |
| 167 | +db.backend=postgres |
| 168 | + |
| 169 | +[postgres] |
| 170 | + |
| 171 | +db.postgres.dsn=postgres://xx:yy@localhost:5432/zz |
| 172 | +``` |
| 173 | + |
| 174 | +Use the same connection string you used for the migration. Also take a look at |
| 175 | +the postgres knobs in the [lnd-sample-config](https://github.com/lightningnetwork/lnd/blob/b6d8ecc7479f7517368814c398b0fbe0e8c52fed/sample-lnd.conf). |
| 176 | + |
| 177 | + |
| 178 | +This is the output of the migration cmd help settings: |
| 179 | + |
| 180 | +```shell |
| 181 | +lndinit migrate-db -h |
| 182 | +2025-04-01 00:29:24.618 [INF]: LNDINIT Version 0.1.24-beta commit=docker/v0.1.22-beta-lnd-v0.18.3-beta.rc1-10-g658c5c8-dirty, debuglevel= |
| 183 | +Usage: |
| 184 | + lndinit [OPTIONS] migrate-db [migrate-db-OPTIONS] |
| 185 | + |
| 186 | +Migrate the full database state of lnd from a source (for example the |
| 187 | +set of bolt database files such as channel.db and wallet.db) database |
| 188 | +to a SQL destination database. |
| 189 | + |
| 190 | +IMPORTANT: Please read the data migration guide located in the file |
| 191 | +docs/data-migration.md of the main lnd repository before using this |
| 192 | +command! |
| 193 | + |
| 194 | +NOTE: The migration can take a long time depending on the amount of data |
| 195 | +that needs to be written! The migration happens in chunks therefore it |
| 196 | +can be resumed in case of an interruption. The migration also includes |
| 197 | +a verification to assure that the migration is consistent. |
| 198 | +As long as NEITHER the source nor destination database has been started/ |
| 199 | +run with lnd, the migration can be repeated/resumed in case of an error |
| 200 | +since the data will just be overwritten again in the destination. |
| 201 | + |
| 202 | +Once a database was successfully and completely migrated from the source |
| 203 | +to the destination, the source will be marked with a 'tombstone' tag |
| 204 | +while the destination will get an 'already migrated' tag. |
| 205 | +A database with a tombstone cannot be started with lnd anymore to |
| 206 | +prevent from an old state being used by accident. |
| 207 | +To prevent overwriting a destination database by accident, the same |
| 208 | +database/namespace pair cannot be used as the target of a data migration |
| 209 | +twice, which is checked through the 'already migrated' tag. |
| 210 | + |
| 211 | +Application Options: |
| 212 | + -e, --error-on-existing Exit with code EXIT_CODE_TARGET_EXISTS (128) instead of 0 if the result of an action is already present |
| 213 | + -d, --debuglevel= Set the log level (Off, Critical, Error, Warn, Info, Debug, Trace) |
| 214 | + |
| 215 | +Help Options: |
| 216 | + -h, --help Show this help message |
| 217 | + |
| 218 | +[migrate-db command options] |
| 219 | + -n, --network= Network of the db files to migrate (used to navigate into the right directory) (default: mainnet) |
| 220 | + --pprof-port= Enable pprof profiling on the specified port |
| 221 | + --force-new-migration Force a new migration from the beginning of the source DB so the resume state will be discarded |
| 222 | + --chunk-size= Chunk size for the migration in bytes |
| 223 | + |
| 224 | + source: |
| 225 | + --source.backend=[bolt] The source database backend. (default: bolt) |
| 226 | + |
| 227 | + bolt: |
| 228 | + --source.bolt.dbtimeout= Specify the timeout value used when opening the database. (default: 1m0s) |
| 229 | + --source.bolt.data-dir= Lnd data dir where bolt dbs are located. |
| 230 | + --source.bolt.tower-dir= Lnd watchtower dir where bolt dbs for the watchtower server are located. |
| 231 | + |
| 232 | + dest: |
| 233 | + --dest.backend=[postgres|sqlite] The destination database backend. (default: postgres) |
| 234 | + |
| 235 | + postgres: |
| 236 | + --dest.postgres.dsn= Database connection string. |
| 237 | + --dest.postgres.timeout= Database connection timeout. Set to zero to disable. |
| 238 | + --dest.postgres.maxconnections= The maximum number of open connections to the database. Set to zero for unlimited. |
| 239 | + |
| 240 | + sqlite: |
| 241 | + --dest.sqlite.data-dir= Lnd data dir where sqlite dbs are located. |
| 242 | + --dest.sqlite.tower-dir= Lnd watchtower dir where sqlite dbs for the watchtower server are located. |
| 243 | + |
| 244 | + sqlite-config: |
| 245 | + --dest.sqlite.sqlite-config.timeout= The time after which a database query should be timed out. |
| 246 | + --dest.sqlite.sqlite-config.busytimeout= The maximum amount of time to wait for a database connection to become available for a query. |
| 247 | + --dest.sqlite.sqlite-config.maxconnections= The maximum number of open connections to the database. Set to zero for unlimited. |
| 248 | + --dest.sqlite.sqlite-config.pragmaoptions= A list of pragma options to set on a database connection. For example, 'auto_vacuum=incremental'. Note that the flag must be |
| 249 | + specified multiple times if multiple options are to be set. |
| 250 | + |
| 251 | + |
| 252 | +``` |
0 commit comments