Skip to content

sptrakesh/config-db

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

88 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Configuration Database

A simple configuration database similar to Apache Zookeeper, Etcd, etc., built using RocksDB.

All values are stored encrypted using aes-256-cbc on disk.

Keys

All keys are assumed to represent UNIX path like structures. Keys that do not at least start with a leading / character are internally replaced with a leading / character. This means that keys key and /key are treated as identical.

The database maintains a hierarchical model of the keys. It is possible to list the child node names under each level in the hierarchy.

Note: When listing child node names, the full path of the parent has to be specified. Relative path names are not supported.

Implementation

Key-value pairs are stored in a RocksDB column family. A few addition column families are used to provide support for additional features provided.

  • data - The main column family used to store the key->value mapping. The value is stored encrypted.
  • forrest - The column family used to store the tree hierarchy.
  • expiration - The column family used to store TTL values for expiring keys.
  • expiring - The column family used to efficiently determine expired keys. The TTL value (along with a value for disambiguation) is stored as the key and the data key is stored as the value in this column. RocksDB keys are stored sorted, hence using the expiration time as the key allows for efficiently breaking from the loop when the first non-expired key is encountered. A custom iterator is used to retrieve only the expired keys from this column.
  • cache - The column family used to indicate that the specified key is stored as a cache entry.

Backup

Daily backups are performed and saved under the storage.backupPath configured. The maximum number of previous backups to maintain can be configured using the storage.maxBackups property. RocksDB makes incremental backups, and the cost of initialising the BackupEngine is proportional to the number of old backups.

Peering

Each instance is designed to run independently. It is possible to configure instances to run as a co-ordinated cluster by configuring peers for each instance. There is no complicated leadership election process/protocol using Raft or similar consensus algorithms. Commands received by each node is relayed to peers. This helps all nodes in the cluster remain in sync with each other eventually.

Commands

The following actions/commands are supported by the service:

  • Put - Set a key-value pair. The key should comply with the above mentioned scheme. The value is an arbitrary length value. Can be large values such as JSON, BSON, base64 encoded binary data etc. as appropriate.
  • Get - Get a stored value for the specified key.
  • Move - Move the value associated with a key to a new destination. Convenience command to perform a Get-Delete-Put in a single transaction.
  • Delete - Delete the key-value pair (and path components as appropriate) identified by the key.
  • List - List child node names under a specified parent path. Leaf nodes will return an error. Note that parent paths are recursively removed during a Delete action if no other child nodes exist.
  • TTL - Retrieve the TTL for the specified key. The returned value indicates the remaining time relative to the request time, not the original value. If the key does not have a TTL policy, 0 is returned.

Protocols

Service supports both TCP/IP and HTTP/2 connections. The TCP service uses flatbuffers as the data interchange format, while the HTTP/2 service uses JSON.

The models were generated from the schema files using the following command:

(cd <path to project>/src/common/model;
<path to>/flatc --cpp --cpp-std c++17 --cpp-static-reflection --reflect-names request.fbs response.fbs tree.fbs)

TCP/IP

The TCP/IP service uses flatbuffers for data interchange. The wire protocol adds an extra 4 bytes at the beginning of each message. These 4 bytes contain the size of the message being interchanged. Client and server will use these 4 bytes to ensure they have read the entire message before attempting to marshall the data.

All requests are for setting or retrieving multiple keys. When setting or deleting multiple keys, a single failure will result in rolling back the entire transaction.

HTTP/2

The HTTP service uses JSON for data interchange. HTTP/1.1 or upgrade to 2 is not supported directly. Clients wishing regular HTTP/1.1 or upgrade feature can use envoy to handle these as well as https.

The HTTP service always responds with a JSON message. Input message is only supported on the PUT endpoints, and only accept a plain text body which is the desired value to set for the key.

The following path prefixes are supported by the service:

  • URI paths with the /key prefix. The rest of the path is interpreted as the desired key within the configuration database. This is used to manage the configuration data.
  • URI paths with the /list prefix. The rest of the path is interpreted as the desired key within the configuration database. This is used to retrieve child node names for the specified path.

See shell and/or cpr for sample HTTP requests and responses.

Custom headers

Request options as defined in the flatbuffers schema can be specified using custom HTTP headers when making request to the HTTP/2 service.

  • x-config-db-if-not-exists - Use to specify the if_not_exists property. A value of true is interpreted as the boolean value true. Any other values are interpreted as false. See the PutNotExists function in the example.
  • x-config-db-ttl - Use to specify the expiration_in_seconds property. Must be a numeric value, and represent the expiration from current time in seconds.
  • x-config-db-cache - Use to specify the cache property. A value of true is interpreted as the boolean value true. Any other values are interpreted as false.

Note: The HTTP service does not support batch (multiple key) operations.

Note: The HTTP service does not support move operations.

PUT

# PUT request
url='http://localhost:6006/key/test'
curl -s --http2-prior-knowledge -XPUT -H "content-type: text/plain" -d "value" $url
{"code": 200, "cause": "Ok"}

PutNotExists

# PUT request
url='http://localhost:6006/key/test'
curl -s --http2-prior-knowledge -XPUT -H "content-type: text/plain" -H "x-config-db-if-not-exists: true" -d "value" $url
{"code": 412, "cause": "Unable to save"}

PutWithTTL

curl -s --http2-prior-knowledge -XPUT \
  -H "content-type: text/plain" \
  -H "x-config-db-if-not-exists: true" \
  -H "x-config-db-ttl: 5" \
  -d "value" 'http://localhost:6006/key/test'
{"code": 200, "cause": "Ok"}

GET

# GET request
url='http://localhost:6006/key/test'
curl -s --http2-prior-knowledge $url
{"key":"/test","value":"value"}

DELETE

# DELETE request
url='http://localhost:6006/key/test'
curl -s --http2-prior-knowledge -XDELETE $url
{"code": 200, "cause": "Ok"}

GET after delete

# GET request after delete
url='http://localhost:6006/key/test'
curl -s --http2-prior-knowledge $url
{"code": 404, "cause": "Not found"}

Messages

The following messages are transferred between the client and TCP server.

Request

The request message contains the action desired as well as the key-value pairs to be sent to the service. The value may be omitted for all actions other than Put.

  • action - An enum used to specify the type of action to perform against the service. Action values are kept similar to their HTTP verb counterparts when possible.
  • data - Array of key-value pairs.
    • key - The key to act upon against the service.
    • value - The value to set for the key. This is only relevant for the Put or Move actions. In the case of Move, the value is the destination key to move the value to.
    • options - Additional options that relate to setting (or moving) keys.
      • if_not_exists - Specify true if key should be saved only if it does not exist in the database. For Move, this applies to the value which serves as the destination key.
      • expiration_in_seconds - TTL in seconds from request time. The key will be automatically deleted by the system after the specified time has elapsed.
      • cache - Indicate that the key-value pair is to be treated as a cache entry. Cached keys are maintained in the tree structure, and are meant to be used purely as temporary values that can be looked up by unique key. Cached keys must have a TTL.

Response

The response message contains either the value for the array of keys (Get, List), or a boolean indicating success or failure of the transaction (Put, Delete).

A KeyValueResult structure is used to represent the result for a specific key sent in the request. The value can be one of:

  • Value - The string value associated with the specified key for a Get request.
  • Children - A list of child node names for the specified key/path for a List request.
  • Success - A boolean value used to report failure at retrieving the key or path.

When reading the response, as mentioned above, the first 4 bytes represent the length of the buffer being returned, and the rest of the bytes (use loops and similar constructs to read until the full message has been received) will be the buffer. See integration test for an example.

Ping

Short (less than 5 bytes) messages may be sent to the service as a keep-alive message. Service will echo the message back to the client. Examples include ping, noop, etc.

API

A high-level client API to interact with the service is provided. The interface hides the complexities involved with making TCP/IP requests using flatbuffers. The api presents an interface that is very similar to the persistence interface used internally by the service. The API maintains a connection pool to the service and performs the required interactions using the flatbuffer models.

Note: API must be initialised via the init function before first use.

See integration test for sample usage of the API. The shell application is built using the client API. See cmake for including the API in your cmake project.

Utilities

Utility applications to interact with the database are provided. These are deployed to the /opt/spt/bin directory.

Shell

The configsh application provides a shell to interact with the database. The server should be running and the TCP port open for the application to connect.

The following command line options are supported by the shell:

  • -s | --server The TCP server hostname to connect to. Default localhost.
  • -p | --port The TCP port to connect to. Default 2020 (2022 on Mac OS X).
  • -t | --with-ssl Flag to connect over SSL to the server.
  • -l | --log-level The level for the logger. Accepted values debug|info|warn|critical. Default info.
  • -o | --log-dir The directory to write log files to. The path must end with a trailing /. Default /tmp/.

The following shows a simple CRUD type interaction via the shell.

Click to expand!
/opt/spt/bin/configsh --server localhost --port 2022 --log-level debug --log-dir /tmp/
Enter commands followed by <ENTER>
Enter help for help about commands
Enter exit or quit to exit shell
configdb> help
Available commands
  ls <path> - To list child node names.  Eg. [ls /]
  get <key> - To get configured value for key.  Eg. [get /key1/key2/key3]
  set <key> <value> - To set value for key.  Eg. [set /key1/key2/key3 Some long value. Note no surrounding quotes]
  mv <key> <destination> - To move value for key to destination.  Eg. [mv /key1/key2/key3 /key/key2/key3]
  rm <key> - To remove configured key.  Eg. [rm /key1/key2/key3]
  import <path to file> - To bulk import key-values from file.  Eg. [import /tmp/kvps.txt]
configdb> set /key1/key2/key3 {"glossary":{"title":"example glossary","GlossDiv":{"title":"S","GlossList":{"GlossEntry":{"ID":"SGML","SortAs":"SGML","GlossTerm":"Standard Generalized Markup Language","Acronym":"SGML","Abbrev":"ISO 8879:1986","GlossDef":{"para":"A meta-markup language, used to create markup languages such as DocBook.","GlossSeeAlso":["GML","XML"]},"GlossSee":"markup"}}}}}
Set key /key1/key2/key3
configdb> ls /
key1
configdb> ls /key1
key2
configdb> ls /key1/key2
key3
configdb> get /key1/key2/key3
{"glossary":{"title":"example glossary","GlossDiv":{"title":"S","GlossList":{"GlossEntry":{"ID":"SGML","SortAs":"SGML","GlossTerm":"Standard Generalized Markup Language","Acronym":"SGML","Abbrev":"ISO 8879:1986","GlossDef":{"para":"A meta-markup language, used to create markup languages such as DocBook.","GlossSeeAlso":["GML","XML"]},"GlossSee":"markup"}}}}}
configdb> set /key1/key2/key3 some long value with spaces and "quoted" text.
Set key /key1/key2/key3
configdb> get /key1/key2/key3
some long value with spaces and "quoted" text.
configdb> rm /key1/key2/key3
Removed key /key1/key2/key3
configdb> ls /
Error retrieving path /
configdb> import /tmp/import.txt
Imported (5/5) keys from file /tmp/import.txt
configdb> exit
Bye

CLI

The configctl application provides a simple means for interacting with the database server. Use it to execute single actions against the service when required.

Note: Unlike the shell application, values specified as command line argument must be quoted if they contain spaces or other special characters.

The server should be running and the TCP port open for the application to connect.

The following command line options are supported by the CLI application:

  • -s | --server The TCP server hostname to connect to. Default localhost.
  • -p | --port The TCP port to connect to. Default 2020 (2022 on Mac OS X).
  • -t | --with-ssl Flag to connect over SSL to the server.
  • -l | --log-level The level for the logger. Accepted values debug|info|warn|critical. Default info.
  • -o | --log-dir The directory to write log files to. The path must end with a trailing /. Default /tmp/.
  • -f | --file The file to bulk import into the database. If specified, other commands are ignored.
  • -a | --action The action to perform. One of get|set|move|delete|list.
  • -k | --key The key to act upon.
  • -v | --value The value to set. For move this is the destination path.

See sample integration test suite.

The following shows a simple CRUD type interaction via the cli. These were using the default values for server [-s|--server] and port [-p|--port] options.

Click to expand!
spt:/home/spt $ /opt/spt/bin/configctl -a list -k /
Error retrieving path /
spt:/home/spt $ /opt/spt/bin/configctl -a set -k /test -v value
Set value for key /test
spt:/home/spt $ /opt/spt/bin/configctl -a list -k /            
test
spt:/home/spt $ /opt/spt/bin/configctl -a set -k /test -v value
Set value for key /test
spt:/home/spt $ /opt/spt/bin/configctl -a set -k /test -v "value modified"                      
Set value for key /test
spt:/home/spt $ /opt/spt/bin/configctl -a get -k /test 
value modified         
spt:/home/spt $ /opt/spt/bin/configctl -a move -k /test -v /test1
Moved key /test to /test1
spt:/home/spt $ /opt/spt/bin/configctl -a delete -k /test1
Removed key /test1
spt:/home/spt $ /opt/spt/bin/configctl -a list -k /
Error listing path /

Bulk Import

A special -f | --file option is supported for bulk importing key-value pairs from a file. The input file must contain lines where each line represents a key-value pair. The text before the first space character is interpreted as the key, and the rest of the line as the value.

/a/b/c string value
/a/b/d 13

Note: The entire file is imported as a single transaction. The input file must be small enough to buffer the requests and transaction in memory.

Click to expand!
spt:/home/spt $ cat /tmp/import.txt
/a/b/c  a long string
/a/b/d another long value
/a/c/e  1234
/a/c/f 45765789
/a/c/g 123.347
spt:/home/spt $ /opt/spt/bin/configctl -f /tmp/import.txt
Set 5 keys

Seed

A seed-configdb utility is provided to seed the database. This can be used to pre-populate the database before the service is started.

Note: This will fail if the service is already running (database locked by running process).

The following command line options are supported by the seed application:

  • -c | --conf Optional JSON configuration file that controls the database storage location.
  • -f | --file The data file to import. The input file must contain lines where each line represents a key-value pair. The text before the first space character is interpreted as the key, and the rest of the line as the value.
  • -l | --log-level The level for the logger. Accepted values debug|info|warn|critical. Default info.
  • -o | --log-dir The directory to write log files to. The path must end with a trailing /. Default /tmp/.

Docker

Docker images are available on Docker hub.

The database files are stored under /opt/spt/data. If you wish to persist the database, volume mount this location.

It is possible to run the integration tests against the docker container instead of the service running locally.

docker run -d --rm -p 6026:6020 -p 2022:2020 --name config-db sptrakesh/config-db
# or with SSL turned on
docker run -d --rm -p 6026:6020 -p 2022:2020 -e "ENABLE_SSL=true" --name config-db sptrakesh/config-db

The following custom environment variables can be used to customise the container (the standard variables also work):

  • CONFIG_FILE - The JSON file (volume mount) with full configuration. All other variables/options are ignored.
  • HTTP_PORT - The port to run the HTTP/2 service on. Default 6020.
  • TCP_PORT - The port to run the TCP service on. Default 2020.
  • NOTIFY_PORT - The port to run the notification service on. Default 2120
  • THREADS - The number of threads to use for the services. Default 4.
  • LOG_LEVEL - The level to use for logging. One of debug|info|warn|critical. Default info.
  • ENABLE_CACHE - Use to turn off temporary value caching. Default true.
  • ENABLE_SSL - Use to run SSL services. Default false.
  • ENCRYPTION_SECRET - Use to specify the secret used to AES encrypt values. Default is internal to the system.
  • PEERS - Use to enable notifications on the NOTIFY_PORT. Listeners will be started to listen for notifications from the PEERS. Will also enable publishing notifications from this instance.

Note: An alpine based image is also available. It crashes at shutdown, so the database may get corrupted.

Build

Standard cmake build procedure. See Dockerfile for build sequence. Ensure the dependencies are available under the following paths:

  • MacOSX - Various dependencies installed under the /usr/local/<dependency> path. See dependencies for scripts used to install the dependencies.
  • UNIX - All dependencies installed under the /opt/local path.
  • Windows - Most dependencies installed under the \opt\local path. A few dependencies also installed via vcpkg under \opt\src\vcpkg.

UNIX

Check out the project and build. Install dependencies. These instructions are for setting up on Mac OS X. For Linux, the assumption is that all dependencies are installed under /opt/local.

git clone [email protected]:sptrakesh/config-db.git
cd config-db
cmake -DCMAKE_PREFIX_PATH=/usr/local/boost \
  -DCMAKE_PREFIX_PATH=/usr/local/rocksdb \
  -DCMAKE_PREFIX_PATH=/usr/local/flatbuffers \
  -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_INSTALL_PREFIX=/usr/local/spt \
  -S . -B build
cmake --build build -j12
sudo cmake --install build

Windows

Install pre-requisites for building and running on Windows. The following instructions at times reference arm or arm64 architecture. Modify those values as appropriate for your hardware. These instructions are based on steps I followed to set up the project on a Windows 11 virtual machine running via Parallels Desktop on a M2 Mac.

Issues

At present final linking of the service fails with unresolved external symbols. These have not manifested themselves on Mac OS X. I will update here if I finally manage to resolve the linking problems.

  • rocksdb::Cleanable::Cleanable(void)
  • rocksdb::PinnableSlice::PinnableSlice(class rocksdb::PinnableSlice &&)
  • rocksdb::Status::ToString(void)const
  • rocksdb::LRUCacheOptions::MakeSharedCache(void)const
  • rocksdb::ColumnFamilyOptions::OptimizeForPointLookup(unsigned __int64)
  • rocksdb::ColumnFamilyOptions::ColumnFamilyOptions(void)
  • rocksdb::DBOptions::OptimizeForSmallDb(class std::shared_ptr<class rocksdb::Cache> *)
  • rocksdb::DBOptions::DBOptions(void)
  • rocksdb::TransactionDB::Open(struct rocksdb::DBOptions const &,struct rocksdb::TransactionDBOptions const &,class std::basic_string<char,struct std::char_traits<char>,class std::allocator<char> > const &,class std::vector<struct rocksdb::ColumnFamilyDescriptor,class std::allocator<struct rocksdb::ColumnFamilyDescriptor> > const &,class std::vector<class rocksdb::ColumnFamilyHandle *,class std::allocator<class rocksdb::ColumnFamilyHandle *> > *,class rocksdb::TransactionDB **)
  • rocksdb::kDefaultColumnFamilyName
Install Boost
  • Download and extract Boost 1.82 (or above) to a temporary location (eg. \opt\src).
  • Launch the Visual Studio Command utility and cd to the temporary location.
cd \opt\src
curl -OL https://boostorg.jfrog.io/artifactory/main/release/1.82.0/source/boost_1_82_0.tar.gz
tar -xfz boost_1_82_0.tar.gz
cd boost_1_82_0
.\bootstrap.bat
.\b2 -j8 install threading=multi address-model=64 architecture=arm asynch-exceptions=on --prefix=\opt\local --without-python --without-mpi

cd ..
del /s /q boost_1_82_0
rmdir /s /q boost_1_82_0
Install fmt library

Launch the Visual Studio Command utility and cd to a temporary location.

cd \opt\src
git clone https://github.com/fmtlib/fmt.git --branch 9.1.0
cd fmt
cmake -DCMAKE_CXX_STANDARD=20 -DCMAKE_CXX_STANDARD_REQUIRED=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=\opt\local -DCMAKE_INSTALL_LIBDIR=lib -DFMT_TEST=OFF -DFMT_MODULE=ON -S . -B build
cmake --build build --target install -j8
Install range-v3 library

Launch the Visual Studio Command utility and cd to a temporary location.

git clone https://github.com/ericniebler/range-v3.git --branch 0.12.0
cd range-v3
cmake -DCMAKE_CXX_STANDARD=20 -DCMAKE_CXX_STANDARD_REQUIRED=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=\opt\local -DCMAKE_INSTALL_LIBDIR=lib -DRANGE_V3_DOCS=OFF -DRANGE_V3_EXAMPLES=OFF -DRANGE_V3_PERF=OFF -DRANGE_V3_TESTS=OFF -DRANGE_V3_INSTALL=ON -B build -S .
cmake --build build --target install -j8
Install flatbuffers
cd \opt\src
git clone -b v23.5.9 https://github.com/google/flatbuffers.git
cd flatbuffers
cmake -DFLATBUFFERS_BUILD_TESTS=OFF -DFLATBUFFERS_BUILD_CPP17=ON -DFLATBUFFERS_ENABLE_PCH=ON -DCMAKE_PREFIX_PATH=\opt\local -DCMAKE_INSTALL_PREFIX=\opt\local -S . -B build
cmake --build build --target install -j8
cd ..
del /s /q flatbuffers
rmdir /s /q flatbuffers
Install vcpkg
cd \opt\src
git clone https://github.com/Microsoft/vcpkg.git
cd vcpkg
.\bootstrap-vcpkg.bat -disableMetrics
.\vcpkg integrate install --vcpkg-root \opt\src\vcpkg
.\vcpkg install openssl:arm64-windows
.\vcpkg install snappy:arm64-windows
.\vcpkg install rocksdb:arm64-windows
.\vcpkg install readline:arm64-windows
Install gflags if compiling RocksDB locally
cd \opt\src
git clone https://github.com/gflags/gflags.git
cd gflags
cmake -DBUILD_SHARED_LIBS=OFF -DBUILD_STATIC_LIBS=ON -DBUILD_TESTING=OFF -DBUILD_gflags_LIBS=ON -DINSTALL_HEADERS=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH=\opt\local -DCMAKE_INSTALL_PREFIX=\opt\local -S . -B cmake-build
cmake --build cmake-build --target install -j8
cd ..
del /s /q gflags
rmdir /s /q gflags
Build RocksDB locally
cd \opt\src
git clone -b v8.3.2 https://github.com/facebook/rocksdb.git
cd rocksdb
cmake -DWITH_TESTS=OFF -DWITH_ALL_TESTS=OFF -DWITH_BENCHMARK_TOOLS=OFF -DCMAKE_CXX_STANDARD=20 -DCMAKE_BUILD_TYPE=Release -DROCKSDB_BUILD_SHARED=OFF -DCMAKE_PREFIX_PATH=\opt\local -DCMAKE_INSTALL_PREFIX=\opt\local -DCMAKE_TOOLCHAIN_FILE="C:/opt/src/vcpkg/scripts/buildsystems/vcpkg.cmake" -S . -B build -G"Unix Makefiles" -DCMAKE_MAKE_PROGRAM=nmake
set CL=/MP
cmake --build build
cd ..
del /s /q rocksdb
rmdir /s /q rocksdb

Note that the current cmake build files do not generate any appropriate install targets, hence the earlier instruction to install via vcpkg.

Build and test the project

Launch the Visual Studio Command utility.

Note: I have not had any success attempting to build nghttp2-asio on Windows ARM64, hence the -DHTTP_SERVER=OFF option is required when invoking cmake.

cd %homepath%\source\repos
git clone https://github.com/sptrakesh/config-db.git
cd config-db
cmake -DCMAKE_PREFIX_PATH=\opt\local -DCMAKE_INSTALL_PREFIX=\opt\spt -DCMAKE_TOOLCHAIN_FILE="C:/opt/src/vcpkg/scripts/buildsystems/vcpkg.cmake" -DHTTP_SERVER=Off -S . -B build
cmake --build build --target install -j8

API Usage

The API can be used to communicate with the TCP service. Client code bases can use cmake to use the library.

# In your CMakeLists.txt file
find_package(ConfigDb REQUIRED COMPONENTS api)
if (APPLE)
  include_directories(/usr/local/spt/include)
else()
  include_directories(/opt/spt/include)
endif (APPLE)
target_link_libraries(${Target_Name} PRIVATE configdb::api ...)

# Run cmake
cmake -DCMAKE_PREFIX_PATH="/usr/local/boost;/usr/local/spt" -S . -B build
cmake --build build -j12

Run

Run the service via the /opt/spt/bin/configdb executable. Command line options may be specified to override default options. When running as a Docker container, use environment variables to specify the command line options.

  • -f | --config-file - The path to the JSON configuration file. All other options are ignored. This option provides total control over all configurable options including encryption. File must have same structure as the struct. See test for sample JSON configuration. For Docker specify the CONFIG_FILE environment variable.
  • Logging - Options related to logging.
    • -c | --console - Flag that controls whether logs are echo'ed to stdout. Default is off. Always specified in the Docker entrypoint. No value (true, false) etc. must be specified.
    • -l | --log-level - The log level to set. Default info. Supported values critical|warn|info|debug. Specify via LOG_LEVEL environment variable to docker.
    • -o | --log-dir - The directory under which log files are written. Default logs/ relative path. On docker this is set to /opt/spt/logs. Files are rotated daily. External scripts (cron etc. are needed to remove old files).
  • -p | --http-port - The port on which the HTTP/2 service listens. Default 6020 (6026 on Apple). Specify via HTTP_PORT environment variable to docker.
  • -t | --tcp-port - The port on which the TCP/IP service listens. Default 2020 (2022 on Apple). Specify via TCP_PORT environment variable to docker.
  • -b | --notify-port - The port on which notifications are published. Default 2120 (2122 on Apple). Specify via NOTIFY_PORT environment variable to docker.
  • -s | --with-ssl - Flag to enable SSL on the HTTP/2 and TCP services.
    See SSL for details. Specify via ENABLE_SSL environment variable to docker.
  • -n | --threads - The number of threads for both TCP/IP and HTTP/2 services. Default to number of hardware threads. Specify via THREADS environment variable to docker.
  • -e | --encryption-secret - The secret to use to encrypt values. Default value is internal to the system. Specify via ENCRYPTION_SECRET environment variable to docker.
  • -x | --enable-cache - Flag to enable temporary cache for keys read from the database. Default off. Specify via ENABLE_CACHE environment variable to docker.
  • -z | --peers - A comma separated list of peer instances to listen for notifications. Specify via the PEERS environment variable to docker. Eg. localhost:2123,localhost:2124

Sample command to run the service

# Locally built service
/opt/spt/bin/configdb --console --log-dir /tmp/ --threads 4 --log-level debug
# Docker container
docker run -d --rm -p 6020:6020 -p 2022:2020 \
  -e "ENCRYPTION_SECRET=svn91sc+rlZXlIXz1ZrGP4m3OgznyW5DrWONGjYw4bc=" -e "LOG_LEVEL=debug" \
  --name config-db config-db

Environment Variables

Service configuration can be customised using environment variables. All properties can be specified using a snake case convention with the common root CONFIG_DB_ prefix.

  • CONFIG_DB_THREADS - use to set the value of the threads field.
  • CONFIG_DB_ENABLE_CACHE - use to set the value of the enableCache field. Set to false to disable (default). Set to true to enable.
  • CONFIG_DB_ENCRYPTION_SALT - use to set the value of the encryption.salt field.
  • CONFIG_DB_ENCRYPTION_KEY - use to set the value of the encryption.key field.
  • CONFIG_DB_ENCRYPTION_IV - use to set the value of the encryption.iv field.
  • CONFIG_DB_ENCRYPTION_SECRET - use to set the value of the encryption.secret field.
  • CONFIG_DB_ENCRYPTION_ROUNDS - use to set the value of the encryption.rounds field.
  • CONFIG_DB_LOGGING_LEVEL - use to set the value of the logging.level field.
  • CONFIG_DB_LOGGING_DIR - use to set the value of the logging.dir field.
  • CONFIG_DB_LOGGING_CONSOLE - use to set the value of the logging.console field.
  • CONFIG_DB_SERVICES_HTTP - use to set the value of the services.http field.
  • CONFIG_DB_SERVICES_TCP - use to set the value of the services.tcp field.
  • CONFIG_DB_SERVICES_NOTIFY - use to set the value of the services.notify field.
  • CONFIG_DB_PEERS[0..6]_HOST - use to set the value of the peers[n].host field (eg. CONFIG_DB_PEERS0_HOST).
  • CONFIG_DB_PEERS[0..6]_PORT - use to set the value of the peers[n].port field.
  • CONFIG_DB_STORAGE_DBPATH - use to set the value of the storage.dbpath field.
  • CONFIG_DB_STORAGE_BACKUP_PATH - use to set the value of the storage.backupPath field.
  • CONFIG_DB_STORAGE_BLOCK_CACHE_SIZE_MB - use to set the value of the storage.blockCacheSizeMb field.
  • CONFIG_DB_STORAGE_CLEAN_EXPIRED_KEYS_INTERVAL - use to set the value of the storage.cleanExpiredKeysInterval field.
  • CONFIG_DB_STORAGE_ENCRYPTER_INITIAL_POOL_SIZE - use to set the value of the storage.encrypterInitialPoolSize field.
  • CONFIG_DB_STORAGE_MAX_BACKUPS - use to set the value of the storage.maxBackups field.
  • CONFIG_DB_STORAGE_USE_MUTEX - use to set the value of the storage.useMutex field.

Notifications

A notification system is available when services are run in a cluster. There is no complicated leadership election process (using Raft or similar) for cluster management/coordination. Each node is designed to run stand-alone. Peer instances are assumed to run independently (multi-master setup). A simple notification system has been implemented which will attempt to keep the independent nodes in sync. Notifications are primarily useful when using a cluster of instances that are also used as a L1/L2 cache on top of application databases. Notifications are sent asynchronously and hence will only achieve eventual consistency in the best case.

When operating in multi-master mode, it is important to configure each instance with the appropriate list of peers. See integration test for sample set up and test suite.

Each instance will publish updates (Put, Delete, and Move) via the notification service. Corresponding listener instances will listen to notifications from the peers.

Notifications are strictly one-way. Notification service only writes messages to the socket, and the listener instances only read messages from the socket. Errors encountered while applying updates on another node do not have any effect on the publishing node.

Notification service sends periodic ping messages to keep the socket connections alive.

SSL

SSL wrappers are enabled for both the TCP and HTTP/2 services. The package includes self-signed certificates. The Makefile can be modified to generate self-signed certificates for your purposes. The easier way to override the certificates is to volume mount the /opt/spt/certs directory when running the docker container.

At present, the file names are hard-coded:

  • ca.crt - The root CA used to verify.
  • server.crt - The server certificate file.
  • server.key - The server key file.
  • client.crt - The client certificate file.
  • client.key - The client key file.

Note: There is definitely an overhead when using SSL. The integration test suite that ran in less than 100ms now takes about 900ms. With 4096 bit keys, it takes about 1.3s.

Note: The TCP service only supports TLS 1.3

Acknowledgements

The following components are used to build this software:

  • RocksDB - The database used to persist configuration.
  • OpenSSL - For AES encryption of the values stored in the database.
  • Boost - In particular Asio for the TCP socket server implementation.
  • nghttp2 - for the HTTP/2 server implementation.
  • flatbuffers - The data interchange format for client-server TCP/IP interactions.
  • nano-signal-slot - Thread safe signal-slot library.
  • concurrentqueue - Lock free concurrent queue implementation for notifications.
  • NanoLog - Logging framework used for the server. I modified the implementation for daily rolling log files.
  • Clara - Command line options parser.
  • Catch2 - Testing framework.
  • cpr - HTTP client library for integration test.