failed to acquire cache lock and failed to unlock

[AIchovy]

Summary

I use kong:cassandra this docker image,and use its default configuration.Then I add a POST method api and make 200 concurrent time post requests continuously.There are some error happend;
This api have 3000 tps when do not use kong, but only 200 now.

It seems that lua-resty-lock this module's problem, i turn up lua_shared_dict cache_locks 100k to 2m.
But still not work.
So,i dont know where the problem is.
Help please...

Steps To Reproduce

1. run images kong:cassandra3&cassandra:2.2
2. add a api
3. use jmeter to post requests

error log:

database_cache.lua:146: get_or_set(): failed to acquire cache lock: timeout, client: x.x.x.x server: kong, request: "POST /producer/stockmsg HTTP/1.1", host: "x.x.x.x:port"
2017/03/07 03:49:01 [error] 112#0: *10925 [lua] database_cache.lua:165: get_or_set(): failed to unlock: unlocked, client: x.x.x.x, server: kong, request: "POST /producer/stockmsg HTTP/1.1", host: "x.x.x.x:port"
...
Constantly appear above the wrong

BTW:i use jmeter to post request.

Additional Details & Logs

• Kong version (0.9.0.0)
• Operating System centos7&docker 1.10.3
• kong docker images kong:cassandra3
• database cassandra 2.2

[Tieske]

• kong docker images kong:cassandra3
• database cassandra 2.2

Which is it? Cassandra 3 or Cassandra 2.2? Note that for Kong 0.9.x only Cassandra 2.x is supported. Kong 0.10 supports Cassandra 3.

Also Kong 0.9.x has seen quite some fixes, so latest for that version is 0.9.9. Have you tried that, or even better the (brand) new 0.10

[Tieske]

see #2186 and #2197

closing this one