Notable configuration options:

When running it for the first time, use the --create option to create the schema and tables.

To wipe the schema and start over, run with --drop --create.

fill-postgresql will start filling the database. It will track real-time updates from nodeos after it catches up.

Use SIGINT or SIGTERM to stop.

Install the following:

• cmake
• A C++17 compiler
• Boost 1.58
• libpqxx
• libpq

Run the following.

git submodule update --init --recursive
mkdir build
cd build
cmake ..
make -j

Run the following on a fresh Ubuntu 18.10 image:

sudo apt update
sudo apt upgrade
sudo apt install build-essential cmake libboost-all-dev git libpq-dev libpqxx-dev
git clone [email protected]:EOSIO/fill-postgresql.git
cd fill-postgresql
git submodule update --init --recursive
mkdir build
cd build
cmake ..
make -j

Caution: either use a firewall to block access to the state-history endpoint, or change it to localhost:8080 to disable remote access.

fill-postgresql relies on postgresql environment variables to establish connections; see the postgresql manual.

A quick-and-dirty way to connect to postgresql server running on another machine is to set these:

• PGUSER
• PGPASSWORD
• PGDATABASE
• PGHOST

Use the psql utility to verify your connection.

This option creates a database which tracks the chain state, but lacks most historical information.

• Get the following:
  • A portable snapshot (data/snapshots/snapshot-xxxxxxx.bin)
  • Optional: a block log which includes the block the snapshot was taken at
• Make sure data/state does not exist
• Start nodeos with the --snapshot option, and the options listed in "Nodeos configuration" above
• Look for Placing initial state in block n in the log
• Start fill-postgresql with --create, --skip-to n, and --trim. Replace n with the value above.
• Do not stop nodeos until it has received at least 1 block from the network, or it won't be able to restart.

If nodeos fails to receive blocks from the network, then try the above using net_api_plugin. Use cleos net disconnect and cleos net connect to reconnect nodes which timed out. Caution when using net_api_plugin: either use a firewall to block access to http-server-address, or change it to localhost:8888 to disable remote access.

Whenever you run fill-postgresql after this point, use the --trim option. Only use --create and --skip-to the first time.

• Enable the producer_api_plugin on a node with full state-history. Caution when using producer_api_plugin: either use a firewall to block access to http-server-address, or change it to localhost:8888 to disable remote access.
• Create a portable snapshot: curl http://127.0.0.1:8888/v1/producer/create_snapshot | json_pp
• Wait for nodeos to process several blocks after the snapshot completed. The goal is for the state-history files to contain at least 1 more block than the portable snapshot has, and for the block log to contain the block after it has become irreversible.
• Note: if the block included in the portable snapshot is forked out, then the snapshot will be invalid. Repeat this process if this happens.
• Stop nodeos
• Make backups of:
  • The newly-created portable snapshot (data/snapshots/snapshot-xxxxxxx.bin)
  • The contents of data/state-history:
    • chain_state_history.log
    • trace_history.log
    • chain_state_history.index: optional. Restoring will take longer without this file.
    • trace_history.index: optional. Restoring will take longer without this file.
  • Optional: the contents of data/blocks, but excluding data/blocks/reversible.

• Get the following:
  • A portable snapshot (data/snapshots/snapshot-xxxxxxx.bin)
  • The contents of data/state-history
  • Optional: a block log which includes the block the snapshot was taken at. Do not include data/blocks/reversible.
• Make sure data/state does not exist
• Start nodeos with the --snapshot option, and the options listed in "Nodeos configuration" above
• Do not stop nodeos until it has received at least 1 block from the network, or it won't be able to restart.

If nodeos fails to receive blocks from the network, then try the above using net_api_plugin. Use cleos net disconnect and cleos net connect to reconnect nodes which timed out. Caution when using net_api_plugin: either use a firewall to block access to http-server-address, or change it to localhost:8888 to disable remote access.