bugfix: clear all cached data when get delete action, and removing stale boolean data in table self.values with a safer way

[membphis]

What this PR does / why we need it:

If the administrator user creates and deletes the same route multiple times, the new route will not be correctly synchronized to the APISIX gateway node, resulting in some request route matching failures and a 404 response.

bugfix: clear all cached data when get delete action, and removing stale boolean data in table self.values with a safer way

it is not easy to add test case. will add later

Pre-submission checklist:

• Did you explain what problem does this PR solve? Or what new features have been added?
• Have you added corresponding test cases?
• Have you modified the corresponding document?
• Is this PR backward compatible?

[membphis]

we'll get the wrong route list in APISIX by this case:

ETCDCTL_API=2 etcdctl rm -r /apisix && make init

make run

curl -i http://127.0.0.1:9080/apisix/admin/routes/1  -H 'X-API-KEY: edd1c9f034335f136f87ad84b625c8f1' -X PUT -d '
{
    "uri": "/route",
    "plugins": {
        "serverless-pre-function": {
            "phase": "rewrite",
            "functions" : ["return function() local router = require(\"apisix.http.router.radixtree_host_uri\")\nlocal routes, routes_ver = router.routes()\n\nlocal core = require(\"apisix.core\")\nngx.say(\"worker_pid: \", ngx.worker.pid())\nngx.say(\"route_ver: \", routes_ver)\nngx.say(\"routes_type: \", type(routes))\n\nlocal keys = {}\nfor key in pairs(routes) do\n    table.insert(keys, key)\nend\n\ntable.sort(keys)\n\nfor _, key in ipairs(keys) do\n    ngx.say(\"--> key: \", key)\n    ngx.say(\"--> val: \", core.json.encode(routes[key], true))\nend\n end"]
        }
    }
}'

curl http://127.0.0.1:9080/route -i

# 200 routes
bash -x create && bash -x delete && bash -x create

curl http://127.0.0.1:9080/route -i

make stop

[membphis]

@apache/apisix-committers please take a look at this PR, I updated the title and reason right now

[marcoabreu]

Please close this PR and lock the conversation. You have just pinged hundreds of Apache people...

[kamaci]

Hi @membphis,

No need to mention apache committers since it notifies all the committers of all ASF projects.

[rmetzger]

to be precise, 3301 people.

[membphis]

Hi @membphis,

No need to mention apache committers since it notifies all the committers of all ASF projects.

sorry for this, I chose wong group. will take care later

[mxm]

Not really @membphis fault. I think this is a configuration issue. Notifying all committers should only be allowed by the admins of the Apache GitHub organization.

[gstein]

@mxm that is not an available configuration option, unfortunately 😥 (believe me: we'd do that in a heartbeat)