LOD: Levels of Detail
An intelligent map tile caching proxy for the edge.
LOD (Levels of Detail) is a thin map tile proxy with in-memory caching and a slim authentication backend. It will sit in front of any tile server and will aggressively cache tiles in memory, optionally storing them in a configured Redis cluster for faster fetching later. LOD is cluster-aware and uses Redis message queueing for intra-cluster communication when multiple instances are deployed together.
LOD is written in Go 1.19 using fiber. TOML is used for configuration. Go templates are used for templating. Internal in-memory caching is built upon the bigcache library by allegro.
Download a build from the releases page or just run:
$ go install github.com/dechristopher/lod@latestNOTE: You'll need the GEOS library installed on your system to use some of LOD's more advanced cache invalidation and priming functionality.
Flags:
--conf Path/URL to TOML configuration file. Default: config.toml
--dev Whether to enable developer mode. Default: false
--debug Optional comma separated debug flags. Ex: foo,bar,baz
--help Shows this help menu.
Usage:
lod [--conf config.toml] [--dev]Or just use our Docker image!
You can create your own Dockerfile that adds a config.toml from the context
into the config directory, like so:
FROM dechristopher/lod:0.9.0
COPY /path/to/your_config.toml /opt/lod_cfg/config.toml
CMD [ "/opt/lod", "--conf", "/opt/lod_cfg/config.toml" ]Alternatively, you can specify something along the same lines with Docker run options:
$ docker run -v /path/to/lod-config:/opt/lod_config -p 1337:1337 lod --conf /opt/lod_config/config.toml- Lightweight, parallel, and non-blocking
- Tileserver agnostic (Martin, Tegola, flat file NGINX, etc.)
- Tile format and content agnostic
- Vector (Mapbox Vector Tiles or other vector formats)
- Raster (PNG/JPG/TIFF)
- And more...
- Supports XYZ (Slippy) and TMS tile indexing schemes
- Multi-level caching
- In-memory, tunable LRU cache as first level
- Redis cluster with configurable TTL as second level
- Dynamic query parameters
- Allow configurable query parameters for tile URLs
- Add to cache key for separate caching (osm/4/5/6/{osm_id})
- Configurable header proxying and deletion
- Configurable headers to pull back into proxied responses from LOD
- Configurable headers to delete from proxied responses from LOD
- Configurable headers to inject into upstream tileserver requests
-
Content-TypeandContent-Encodingadded by default
- Internal stats tracking
- Hits, misses, hit-rate
- Tiles per second (load averages)
- Tile upstream fetch times (avg, 75th, 99th)
- Expose Prometheus endpoint
- Supports multiple configured tileserver proxies
- Separate authentication (bearer tokens and CORS)
- Separate internal cache instances per proxy
- Separate stats tracking
- Administrative endpoints
- Security via Bearer Token Authorization
- Reload the instance configuration
- Flush the instance caches
- Invalidate a given tile and re-prime it
- Iteratively invalidate all tiles under a given tile (all zoom levels)
- Iteratively prime all tiles under a given tile
- Cluster-wide operations
- Flush the instance caches across all instances
- Invalidate a given tile and re-prime it across the cluster
A more verbose version of this config actually used for internal testing can be found at config.toml.example in the root of the repo.
More detailed information about configuring LOD and hardening it for production use can be found by reading the source code.
[instance]
# port to bind to
port = 1337
# admin endpoint bearer token
admin_token = "${ADMIN_TOKEN}" # config supports environment variables
# base proxy configuration
[[proxies]]
# name of this proxy, available at http://lod/{name}/{z}/{x}/{y}.{file_extension}
name = "osm"
# url of the upstream tileserver with template parameters
# for the X, Y, and Z values. These are required.
tile_url = "https://tile.example.com/osm/{z}/{x}/{y}.pbf"
# comma-separated list of allowed CORS origins
cors_origins = "https://example.com"
# auth token (?token=XXX) to require for requests to upstream tileserver
access_token = "MyTilesArePrivate"
# headers to pull and cache from the tileserver response
pull_headers = ["X-We-Want-This", "X-This-One-Too"]
# headers to delete from the tileserver response
del_headers = ["X-Get-Rid-Of-Me"]
# proxy cache configuration
[proxies.cache]
# enable in-memory cache
mem_enabled = true
# maximum capacity in MB of the in-memory cache
mem_cap = 100
# Cache TTLs are set using Go's built-in time.ParseDuration
# Valid time units are "ns", "us" (or "µs"), "ms", "s", "m", "h".
# For example: 1h, 5m, 300s, 1000ms, 2h35m, etc.
# in-memory cache TTL
mem_ttl = "1h"
# enable redis cache
redis_enabled = true
# redis tile cache TTL, or "0" for no expiry
redis_ttl = "24h"
# redis connection URL
redis_url = "redis://localhost:6379/0"
# cache key template string, supports parameter names
key_template = "{z}/{x}/{y}"
# headers to inject into upstream tileserver requests
[[proxies.add_headers]]
# name of header to add
name = "Referer"
# value of header to add
value = "https://yoursite.com/"
# Supports many configured proxy instances for caching multiple tileservers
[[proxies]]
name = "another"
# etc.WARNING: these are experimental configuration properties. Only change them if you know what you're doing. All are exposed as environment variables.
MAX_ENTRY_SIZE: int
Size in MB of the "entry" that bigcache sizes its internal cache buckets by. Should be about the size of your 90th
percentile tiles.
LOD is licensed under the GNU Affero General Public License 3 or any later version at your choice. See COPYING for details.