18 lines
1.1 KiB
Markdown
18 lines
1.1 KiB
Markdown
#wp-metacache
|
|
|
|
This is a combination wordpress plugin and Go service that replaces Wordpress' built in object metadata handling.
|
|
|
|
Wordpress has two types of object, posts and users. Both have metadata tables and are accessed through the same mechanisms.
|
|
|
|
Wordpress uses a greedy load system where loading a single metadata attribute will actually load all metadata for that object,
|
|
and cache the other entries for the life of the request. This more or less works well on small sites where there may be a few
|
|
or a dozen metadata attributes, but doesn't scale well as those numbers increase. At the same time, the lifespan of that cache
|
|
is a single request.
|
|
|
|
This software uses Wordpress hooks to handle those metadata requests via an external Go process. In Go we implement a LRU cache
|
|
to greedily load metadata but return it a single item at a time. The Go process will retain the data until we reach a configured
|
|
max capacity, at which point the longest unused entry will be removed. We also hook into Wordpress metadata update and delete
|
|
routines to explicitly remove entries when stale.
|
|
|
|
Communication with the Go process is via a Unix socket.
|