To purge local cache need to restart apisix?

this is the core reason whey the new way is better performance.

use the cached data, reduce to recall secret.fetch_secrets

I think we can use lrucache, which is easier way, here is the code example:

local lrucache = core.lrucache.new({
    ttl = 300, count = 5000  -- we can change the `count` as we need, in your case, we may change it to more than 17,000+
})

local new_consumer, err = lrucache(consumer, nil,
                    secret.fetch_secrets, consumer.auth_conf, false)

I can not 100% confirm if it works, welcome to make a test in this way, many thx

this is the core reason whey the new way is better performance.

use the cached data, reduce to recall secret.fetch_secrets

I think we can use `lrucache`, which is easier way, here is the code example:

local lrucache = core.lrucache.new({
    ttl = 300, count = 5000  -- we can change the `count` as we need, in your case, we may change it to more than 17,000+
})

local new_consumer, err = lrucache(consumer, nil,
                    secret.fetch_secrets, consumer.auth_conf, false)

I can not 100% confirm if it works, welcome to make a test in this way, many thx

Let me make the ci happy first, and then I'll try your suggestion.

ping @membphis

To purge local cache need to restart apisix?

yes

The code introduces a variable consumers_count_for_lrucache, which is used to set the size of the LRU cache. In actual use, this value needs to be configured to be greater than the actual number of consumers. Additionally, if a consumer has multiple authentication plugins, each of them will be considered as a separate consumer and cached individually in the LRU cache.

When the number of consumers exceeds twice the size of the LRU cache, the process of creating cache during the execution of the consumer.plugin function may result in cache misses, leading to performance degradation. This happens because the LRU cache can only store consumers within a single window, and each time a consumer is updated, the order of the consumer IDs in the consumers.values array is likely to either remain the same or change very little. As a result, new consumer objects are continuously created, causing the old consumer cache entries to be evicted.

代码中引入了一个变量 consumers_count_for_lrucache , 用来设置 lrucache 的大小。实际使用的时候，这个值需要配置为大于 consumer 的实际数量，并且如果一个consumer 中存在多个认证插件，则会被认为是不同的consumer 缓存到 lrucache 中。

当 consumer 数量大于2倍的lrucache 大小的时候，在执行 consumer.plugin 函数创建缓存的过程中可能会一直不能命中缓存而导致性能下降。因为 lrucache 只能缓存一个窗口内的 consumer，而每次 consumer 有更新时，consumers.values 数组中的 consumer id 的顺序大概率可能不变或者变化很小，此时会一直存在生成新的 consumer object ，淘汰旧的 consumer 缓存。

ping @membphis

many thx @xuruidong 's contribution, merged your PR right now