using nginx + lua + redis for redirects and rewrites

100_base.conf

# using such a setup requires `apt-get install lua-nginx-redis` under Ubuntu Trusty
# more info @ http://wiki.nginx.org/HttpLuaModule#access_by_lua

http {
  lua_package_path "/etc/nginx/include.d/?.lua;;";
  lua_socket_pool_size 100;
  lua_socket_connect_timeout 10ms;
  lua_socket_read_timeout 10ms;
  
  server {
    listen       80;
    server_name  www.stylight.com;

    location / {
      include /etc/nginx/include.d/en_US/rewrites/122_lua_st_redis_rewrites.lua;
    }
  }
}

122_lua_st_redis_rewrites.lua

access_by_lua '
  local redis = require "redis"
  local red = redis:new()
  -- see all available redis: methods @ https://github.com/openresty/lua-resty-redis#methods
  local ok, err = red:connect("127.0.0.1", 6379)
  if not ok then
    ngx.log(ngx.ERR, "failed to connect to redis: ", err)
    return ngx.exit(500)
  end
  red:select(1)
  local redirect_candidate = ngx.var.uri
  local args_remainder = ngx.null
  local args = ngx.var.args

  -- we store internal application rewrites as hash values in Redis for their corresponding pretty URL keys
  -- if a bot crawls through asking for /search.action?brand=3210
  local hash_rewrite, error = red:hgetall(ngx.var.uri)
  -- 127.0.0.1:6379[1]> hgetall /search.action?brand=3210
  -- (error) WRONGTYPE Operation against a key holding the wrong kind of value
  -- let's find the appropriate redirect URL:
  if error then
    if args then
      local m = ngx.re.match(args, [[^(gclid=.*)?(?<start>.*(shop|brand|tag)=[0-9]+)+&?(.*(shop|sale|searchstring|gender)=.*?&?)*(|&(?<remainder>.*))$]])
      if m then
        if m["start"] then
          redirect_candidate = redirect_candidate .. "?" .. m["start"]
        end
        if m["remainder"] then
          args_remainder = m["remainder"]
        end
      end
    end

    -- and ask Redis for the 301 redirect URL
    local rewrite_value, err = red:get(redirect_candidate)
    -- 127.0.0.1:6379[1]> get "/search.action?brand=3210"
    -- "/Mexx/"

    -- create/use a connection pool of size 100 with 0 idle timeout
    red:set_keepalive(0, 100)
    if rewrite_value ~= ngx.null then
      if args_remainder ~= ngx.null then
        rewrite_value = rewrite_value .. "?" .. args_remainder
      end
      -- and kindly ask the bot to visit /Mexx/ instead
      ngx.redirect(rewrite_value, 301);
    end
  else
    -- a request for /Sandals/Women/ however
    if table.getn(hash_rewrite) > 0 then
      red:set_keepalive(0, 100)
      
      -- gives us the following hash from Redis:
      -- 127.0.0.1:6379[3]> hgetall /Sandals/Women/
      -- 1) "action"
      -- 2) "search.action"
      -- 3) "params"
      -- 4) "gender=women&tag=10580&tag=10630"
      
      -- lets transform to an internal rewrite
      arr_rewrite = red:array_to_hash(hash_rewrite)
      params = nil
      if arr_rewrite["params"] then
        params = arr_rewrite["params"]
      end
      if args then
        if params~=nil then
          params = params .. "&" .. args
        else
          params = args
        end
      end
      if params~=nil then
        ngx.req.set_uri_args(params)
      end
      -- and send off to our app server for processing
      ngx.req.set_uri("/" .. arr_rewrite["action"])
    end
  end
';

Note that this requires the Nginx Lua module, so Nginx must be compiled with it or it must dynamically load the Lua module. I've used the fabiocicerchia/nginx-lua Docker image, as well as Kubernetes' ingress-nginx.

I'd recommend using access_by_lua_file instead of access_by_lua so you can store the Lua script in its own pure-Lua file. This helps with syntax highlighting and any code quality tools you might use.

very valid points @roguh!

for context (since timestamps are a bit obfuscated on gists), I wrote/used this script back in 2014