[Help Wanted] High concurrency performance issues

[JobberRT]

Hi everyone, first of all thanks for reading and replying this issue!

I got a application, which basically can be described as:

• state API, which is requested at a low frequency(only when needed)
• get API, which is requested at a very high frequency(about 100~500 clients, each client requests about 5times/s)

And here's the issues:
After echo started, if get haven't been requested or only requested in a low frequency(such as when i debug, i shutdown most of the client), the state still works and has a low response latency.
But once get start to been requested at a high frequency, the state will not be responsed and chrome will not show Error request timeout message. Even i lower the frequency to about 5 req/s in total, state will not be responsed again.

So what cause the problem and what should i solve this?

I've tried add lock for these two apis, when state got requested, it locks, and get will be blocked.But this will cause huge response latency for the clients, I don't think this is a good solution.

[JobberRT]

In additional to this, I found that the get api will not be responsed after a high frequency requestion peroid.So in total: Once an api got high frequency requesting peroid, even lower the frequency, all apis will not be responsed again

[JobberRT]

It turns out that it should be a sync.Mutex problem.

There's a lock sLock, and every get request will lock and unlock sLock, and state will wait for sLock to be unlocked.
It should be request not been process fast enough, causing sLock keeps locked for a while after shutdown all the client, but I also found another issue:

I log the time cost during middleware and echo's json encode and all the functions which be executed by get api, it shows that my total cost is around 1~5ms, but my debug client reports it use around 100ms, client is in same pc(localhost)