passive healthcheck not working after configuration changing #4077

Ben0625 · 2021-04-19T03:24:55Z

Issue description

When both active and passive configuration are updated from only active configuration, the http statuses of passive become an empty table which causes passive healthcheck not working.

Environment

apisix version: 2.0
OS: 3.10.0-693.el7.x86_64
OpenResty version: openresty/1.17.8.2
etcd version: 3.4.13
apisix-dashboard version: 2.0

Minimal test code / Steps to reproduce the issue

Set a route with two upstreams(/test.do) and healthcheck active configuration(/check.do).
Send a request to trigger creating new checker.
Update this route only by changing healthcheck configuration to both active and passive. Let one of the upstream return 500 code to act like a unnormal node. Send some requests, but the unnormal node cannot be set to be unhealthy even reaching the passive.unhealthy.http_failures.

What's the actual result?

I prints some debug info inside this function: https://github.com/Kong/lua-resty-healthcheck/blob/master/lib/resty/healthcheck.lua#L1372
I prints out the "defaults" table at line 1443.

After doing Step 2, the type of defaults.checks.passive.unhealthy.http_statuses changes from array to dict!

defaults.checks.passive.unhealthy.http_statuses before to_set() function:
{
1 = 429,
2 = 500,
3 = 503,
}

defaults.checks.passive.unhealthy.http_statuses after to_set() function:
{
429 = true,
500 = true,
503 = true,
}

It seems that to_set() changes the "self" object, result in changing "defaults" as well.

What's the expected result?

In my opinion, the "defaults" object should not be changed. Maybe there's some problem with fill_in_settings() function at line 1377. Or am I doing something wrong during these operations?

Firstsawyou · 2021-04-19T04:21:49Z

You should provide the corresponding health check configuration information so that we can check it.

Ben0625 · 2021-04-19T06:25:54Z

For Step 1:

"checks": {
			"active": {
				"healthy": {
					"interval": 100,
					"successes": 4
				},
				"host": "aaa.com",
				"http_path": "/check.do",
				"timeout": 3,
				"unhealthy": {
					"http_failures": 5,
					"interval": 3
				}
			}
		}

For Step 3:

"checks": {
		"active": {
			"healthy": {
				"interval": 100,
				"successes": 4
			},
			"host": "aaa.com",
			"http_path": "/check.do",
			"timeout": 3,
			"unhealthy": {
				"http_failures": 5,
				"interval": 3
			}
		},
		"passive": {
			"healthy": {
				"http_statuses": [200],
				"successes": 1
			},
			"unhealthy": {
				"http_failures": 1,
				"http_statuses": [500],
				"tcp_failures": 1
			}
		}
	}

Firstsawyou · 2021-04-19T11:29:15Z

I tested it, by looking at the log output, and it worked.

tokers · 2021-04-19T12:50:59Z

@Ben0625 Could you try the 2.5 version? And test it whether it can be reproduced in 2.5?

Ben0625 · 2021-04-20T07:34:35Z

@Ben0625 Could you try the 2.5 version? And test it whether it can be reproduced in 2.5?

I tried the Apisix 2.5 version but not use the dashboard 2.5 version. I just connect to the same etcd and it can be reproduced. Let me provide more details.
Step 1. Set a route with two upstreams(/test.do) and healthcheck active configuration(/check.do) through dashboard.
Step 2. Send a request to trigger creating new checker. By looking at the log output before and after to_set() function:

defaults.checks.passive.unhealthy.http_statuses before to_set() function:
{
1 = 429,
2 = 500,
3 = 503,
}
defaults.checks.passive.unhealthy.http_statuses after to_set() function:
{
429 = true,
500 = true,
503 = true,
}

Step 3. Edit the same route on dashboard and do not change any configuration and commit.
Step 4. Send a request to trigger creating new checker. By looking at the log output before and after to_set() function:

defaults.checks.passive.unhealthy.http_statuses before to_set() function:
{
429 = true,
500 = true,
503 = true,
}
defaults.checks.passive.unhealthy.http_statuses after to_set() function:
{
}

The configuration in etcd(/apisix/routes/350659784246034728):
{"id":"350659784246034728","create_time":1618539118,"update_time":1618902032,"uris":["/test.do"],"name":"test","methods":["GET","HEAD","POST","PUT","DELETE","OPTIONS","PATCH"],"vars":[],"upstream":{"nodes":[{"host":"172.25.69.182","port":8080,"weight":1},{"host":"172.25.70.116","port":8080,"weight":1}],"timeout":{"connect":6000,"read":6000,"send":6000},"type":"roundrobin","checks":{"active":{"healthy":{"interval":100,"successes":4},"host":"aaa.com","http_path":"/check.do","timeout":3,"unhealthy":{"http_failures":5,"interval":3}}}}}

tzssangglass · 2021-04-29T13:08:46Z

@Ben0625 any news? Sure this problem still exists in apisix 2.5?

chzhuo · 2021-11-10T08:27:57Z

@Ben0625 Could you try the 2.5 version? And test it whether it can be reproduced in 2.5?

We get the same problems on the apisix 2.8

I found a bug in https://github.com/api7/lua-resty-healthcheck/blob/master/lib/resty/healthcheck.lua
The content of var defaults will be modified when not assigned the passive or active config item

Reproduce the bug:
nginx.conf

daemon off;
worker_processes  1;
error_log stderr;
events {
    worker_connections 1024;
}
http {
    access_log off;
    server {
        listen 19000;
        location / {
            default_type text/html;
            content_by_lua '
                ngx.say("<p>Hello, World!</p>")
            ';
            log_by_lua '
                local check = require("check")
            ';
        }
    }
}

check.lua

local cjson = require("cjson.safe")

-- copy from: https://github.com/api7/lua-resty-healthcheck/blob/master/lib/resty/healthcheck.lua#L1162
--============================================================================
-- Create health-checkers
--============================================================================


local NO_DEFAULT = {}
local MAXNUM = 2^31 - 1


local function fail(ctx, k, msg)
  ctx[#ctx + 1] = k
  error(table.concat(ctx, ".") .. ": " .. msg, #ctx + 1)
end


local function fill_in_settings(opts, defaults, ctx)
  ctx = ctx or {}
  local obj = {}
  for k, default in pairs(defaults) do
    local v = opts[k]

    -- basic type-check of configuration
    if default ~= NO_DEFAULT
    and v ~= nil
    and type(v) ~= type(default) then
      fail(ctx, k, "invalid value")
    end

    if v ~= nil then
      if type(v) == "table" then
        if default[1] then -- do not recurse on arrays
          obj[k] = v
        else
          ctx[#ctx + 1] = k
          obj[k] = fill_in_settings(v, default, ctx)
          ctx[#ctx + 1] = nil
        end
      else
        if type(v) == "number" and (v < 0 or v > MAXNUM) then
          fail(ctx, k, "must be between 0 and " .. MAXNUM)
        end
        obj[k] = v
      end
    elseif default ~= NO_DEFAULT then
      obj[k] = default
    end

  end
  return obj
end


local defaults = {
  name = NO_DEFAULT,
  shm_name = NO_DEFAULT,
  type = NO_DEFAULT,
  status_ver = 0,
  checks = {
    active = {
      type = "http",
      timeout = 1,
      concurrency = 10,
      http_path = "/",
      https_verify_certificate = true,
      healthy = {
        interval = 0, -- 0 = disabled by default
        http_statuses = { 200, 302 },
        successes = 2,
      },
      unhealthy = {
        interval = 0, -- 0 = disabled by default
        http_statuses = { 429, 404,
                          500, 501, 502, 503, 504, 505 },
        tcp_failures = 2,
        timeouts = 3,
        http_failures = 5,
      },
      req_headers = {""},
    },
    passive = {
      type = "http",
      healthy = {
        http_statuses = { 200, 201, 202, 203, 204, 205, 206, 207, 208, 226,
                          300, 301, 302, 303, 304, 305, 306, 307, 308 },
        successes = 5,
      },
      unhealthy = {
        http_statuses = { 429, 500, 503 },
        tcp_failures = 2,
        timeouts = 7,
        http_failures = 5,
      },
    },
  },
}


local function to_set(tbl, key)
  local set = {}
  for _, item in ipairs(tbl[key]) do
    set[item] = true
  end
  tbl[key] = set
end


  -- In the below is my code
local option_no_passive  =  cjson.decode([[
    {
        "checks":
        {
            "active":{"unhealthy":{"tcp_Failures":3,"http_statuses":[500,501,502,503,504,505],"timeouts":3,"interval":100,"tcp_failures":2,"http_failures":3},"concurrency":10,"http_path":"\/healthz","https_verify_certificate":true,"port":9080,"type":"http","healthy":{"interval":5,"successes":1,"http_statuses":[200,201,202,203,204,205,206,207,208,226,300,301,302,303,304,305,306,307,308,400,401,403,404,405,415,429]},"host":"healthcheck","timeout":5}
        },
        "name":"upstream#\/internal-apisix\/upstreams\/380542682478412600","shm_name":"upstream-healthcheck"
    }]])
-- The default can be encoded to json
ngx.log(ngx.ERR, "===defaults: ", cjson.encode(defaults))

-- The default `passive` config will be filled into `filled_option_no_passive` by the method `fill_in_settings`
local filled_option_no_passive = fill_in_settings(option_no_passive ,defaults)
ngx.log(ngx.ERR, "===filled_option_no_passive: ", cjson.encode(filled_option_no_passive))

-- copy from: https://github.com/api7/lua-resty-healthcheck/blob/master/lib/resty/healthcheck.lua#L1364
to_set(filled_option_no_passive.checks.active.unhealthy, "http_statuses")
to_set(filled_option_no_passive.checks.active.healthy, "http_statuses")
to_set(filled_option_no_passive.checks.passive.unhealthy, "http_statuses")
to_set(filled_option_no_passive.checks.passive.healthy, "http_statuses")

-- print: nil Cannot serialise table: excessively sparse array while logging request
-- The defaults has been modified by `to_set`
ngx.log(ngx.ERR, "===defaults: ", cjson.encode(defaults))

shuaijinchao · 2021-11-11T10:25:07Z

@Ben0625 Could you try the 2.5 version? And test it whether it can be reproduced in 2.5?

I tried the Apisix 2.5 version but not use the dashboard 2.5 version. I just connect to the same etcd and it can be reproduced. Let me provide more details. Step 1. Set a route with two upstreams(/test.do) and healthcheck active configuration(/check.do) through dashboard. Step 2. Send a request to trigger creating new checker. By looking at the log output before and after to_set() function:

defaults.checks.passive.unhealthy.http_statuses before to_set() function:
{
1 = 429,
2 = 500,
3 = 503,
}
defaults.checks.passive.unhealthy.http_statuses after to_set() function:
{
429 = true,
500 = true,
503 = true,
}

Step 3. Edit the same route on dashboard and do not change any configuration and commit. Step 4. Send a request to trigger creating new checker. By looking at the log output before and after to_set() function:

defaults.checks.passive.unhealthy.http_statuses before to_set() function:
{
429 = true,
500 = true,
503 = true,
}
defaults.checks.passive.unhealthy.http_statuses after to_set() function:
{
}

The configuration in etcd(/apisix/routes/350659784246034728): {"id":"350659784246034728","create_time":1618539118,"update_time":1618902032,"uris":["/test.do"],"name":"test","methods":["GET","HEAD","POST","PUT","DELETE","OPTIONS","PATCH"],"vars":[],"upstream":{"nodes":[{"host":"172.25.69.182","port":8080,"weight":1},{"host":"172.25.70.116","port":8080,"weight":1}],"timeout":{"connect":6000,"read":6000,"send":6000},"type":"roundrobin","checks":{"active":{"healthy":{"interval":100,"successes":4},"host":"aaa.com","http_path":"/check.do","timeout":3,"unhealthy":{"http_failures":5,"interval":3}}}}}

Following the steps you provided, I tested it through the CLI and Dashboard in the APISIX2.8 version, and this problem was not reproduced.

shuaijinchao · 2021-11-11T10:39:39Z

@Ben0625 Could you try the 2.5 version? And test it whether it can be reproduced in 2.5?

We get the same problems on the apisix 2.8

I found a bug in api7/lua-resty-healthcheck@master/lib/resty/healthcheck.lua The content of var defaults will be modified when not assigned the passive or active config item

Reproduce the bug: nginx.conf

daemon off;
worker_processes  1;
error_log stderr;
events {
    worker_connections 1024;
}
http {
    access_log off;
    server {
        listen 19000;
        location / {
            default_type text/html;
            content_by_lua '
                ngx.say("<p>Hello, World!</p>")
            ';
            log_by_lua '
                local check = require("check")
            ';
        }
    }
}

check.lua

local cjson = require("cjson.safe")

-- copy from: https://github.com/api7/lua-resty-healthcheck/blob/master/lib/resty/healthcheck.lua#L1162
--============================================================================
-- Create health-checkers
--============================================================================


local NO_DEFAULT = {}
local MAXNUM = 2^31 - 1


local function fail(ctx, k, msg)
  ctx[#ctx + 1] = k
  error(table.concat(ctx, ".") .. ": " .. msg, #ctx + 1)
end


local function fill_in_settings(opts, defaults, ctx)
  ctx = ctx or {}
  local obj = {}
  for k, default in pairs(defaults) do
    local v = opts[k]

    -- basic type-check of configuration
    if default ~= NO_DEFAULT
    and v ~= nil
    and type(v) ~= type(default) then
      fail(ctx, k, "invalid value")
    end

    if v ~= nil then
      if type(v) == "table" then
        if default[1] then -- do not recurse on arrays
          obj[k] = v
        else
          ctx[#ctx + 1] = k
          obj[k] = fill_in_settings(v, default, ctx)
          ctx[#ctx + 1] = nil
        end
      else
        if type(v) == "number" and (v < 0 or v > MAXNUM) then
          fail(ctx, k, "must be between 0 and " .. MAXNUM)
        end
        obj[k] = v
      end
    elseif default ~= NO_DEFAULT then
      obj[k] = default
    end

  end
  return obj
end


local defaults = {
  name = NO_DEFAULT,
  shm_name = NO_DEFAULT,
  type = NO_DEFAULT,
  status_ver = 0,
  checks = {
    active = {
      type = "http",
      timeout = 1,
      concurrency = 10,
      http_path = "/",
      https_verify_certificate = true,
      healthy = {
        interval = 0, -- 0 = disabled by default
        http_statuses = { 200, 302 },
        successes = 2,
      },
      unhealthy = {
        interval = 0, -- 0 = disabled by default
        http_statuses = { 429, 404,
                          500, 501, 502, 503, 504, 505 },
        tcp_failures = 2,
        timeouts = 3,
        http_failures = 5,
      },
      req_headers = {""},
    },
    passive = {
      type = "http",
      healthy = {
        http_statuses = { 200, 201, 202, 203, 204, 205, 206, 207, 208, 226,
                          300, 301, 302, 303, 304, 305, 306, 307, 308 },
        successes = 5,
      },
      unhealthy = {
        http_statuses = { 429, 500, 503 },
        tcp_failures = 2,
        timeouts = 7,
        http_failures = 5,
      },
    },
  },
}


local function to_set(tbl, key)
  local set = {}
  for _, item in ipairs(tbl[key]) do
    set[item] = true
  end
  tbl[key] = set
end


  -- In the below is my code
local option_no_passive  =  cjson.decode([[
    {
        "checks":
        {
            "active":{"unhealthy":{"tcp_Failures":3,"http_statuses":[500,501,502,503,504,505],"timeouts":3,"interval":100,"tcp_failures":2,"http_failures":3},"concurrency":10,"http_path":"\/healthz","https_verify_certificate":true,"port":9080,"type":"http","healthy":{"interval":5,"successes":1,"http_statuses":[200,201,202,203,204,205,206,207,208,226,300,301,302,303,304,305,306,307,308,400,401,403,404,405,415,429]},"host":"healthcheck","timeout":5}
        },
        "name":"upstream#\/internal-apisix\/upstreams\/380542682478412600","shm_name":"upstream-healthcheck"
    }]])
-- The default can be encoded to json
ngx.log(ngx.ERR, "===defaults: ", cjson.encode(defaults))

-- The default `passive` config will be filled into `filled_option_no_passive` by the method `fill_in_settings`
local filled_option_no_passive = fill_in_settings(option_no_passive ,defaults)
ngx.log(ngx.ERR, "===filled_option_no_passive: ", cjson.encode(filled_option_no_passive))

-- copy from: https://github.com/api7/lua-resty-healthcheck/blob/master/lib/resty/healthcheck.lua#L1364
to_set(filled_option_no_passive.checks.active.unhealthy, "http_statuses")
to_set(filled_option_no_passive.checks.active.healthy, "http_statuses")
to_set(filled_option_no_passive.checks.passive.unhealthy, "http_statuses")
to_set(filled_option_no_passive.checks.passive.healthy, "http_statuses")

-- print: nil Cannot serialise table: excessively sparse array while logging request
-- The defaults has been modified by `to_set`
ngx.log(ngx.ERR, "===defaults: ", cjson.encode(defaults))

@chzhuo your question does not seem to be related to @Ben0625's question. Regarding the issue of "Unable to serialize table: too sparse array", please refer to the lua-cjson document: kyne.com.au/~mark/software/lua-cjson -manual.html#encode_sparse_array

chzhuo · 2021-11-12T03:30:21Z

@shuaijinchao In you last screenshot, do you found out that the content of the global variable defaults has been modifed?

This change will effect the other upstream to call healthcheck.new()

chzhuo · 2021-11-12T04:19:32Z

@shuaijinchao You can reproduce in apisix by following step:

add log(self) in https://github.com/api7/lua-resty-healthcheck/blob/master/lib/resty/healthcheck.lua#L1326
add the first upstream with only active config
add the second upstream with active and passive config
assign route to the two upstream
curl the route that assign to first upstream
curl the route that assign to second upstream
checking the passive http_statues from log(self) is correct

notice: apisix won't start the health check when the upstream only has one node

shuaijinchao · 2021-11-14T14:51:46Z

@chzhuo Thanks, I reproduced the problem. The reason for this problem is that when healthcheck fills the default value, the table object in the defaults variable is directly assigned to the opts variable, while the shallow copy used in the table type variable assignment in Lua. So when using to_set to update the opts variable, the defaults variable will also be affected, because the memory address of the operation is the same.

shuaijinchao · 2021-11-14T14:57:28Z

At present, this problem still exists in the latest version of APISIX, and we will fix this problem in the recent version.

This issue is introduced in the bugfix: apache#5589 This PR fixes the bug added by that bugfix, and also introduce a new test (TEST 6) to cover issue apache#4077. The fix of that issue is submitted in api7/lua-resty-healthcheck#5 which fixes the bug in an appropriate way. Signed-off-by: spacewander <spacewanderlzx@gmail.com>

shuaijinchao added the bug Something isn't working label Nov 14, 2021

spacewander assigned shuaijinchao Nov 19, 2021

shuaijinchao mentioned this issue Nov 23, 2021

fix: invalid error after passive health check is changed #5589

Merged

4 tasks

spacewander closed this as completed in #5589 Nov 24, 2021

spacewander mentioned this issue Sep 2, 2022

fix: don't enable passive healthcheck by default #7850

Merged

5 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

passive healthcheck not working after configuration changing #4077

passive healthcheck not working after configuration changing #4077

Ben0625 commented Apr 19, 2021 •

edited

Loading

Firstsawyou commented Apr 19, 2021

Ben0625 commented Apr 19, 2021 •

edited

Loading

Firstsawyou commented Apr 19, 2021

tokers commented Apr 19, 2021

Ben0625 commented Apr 20, 2021 •

edited

Loading

tzssangglass commented Apr 29, 2021

chzhuo commented Nov 10, 2021

shuaijinchao commented Nov 11, 2021

shuaijinchao commented Nov 11, 2021 •

edited

Loading

chzhuo commented Nov 12, 2021

chzhuo commented Nov 12, 2021

shuaijinchao commented Nov 14, 2021

shuaijinchao commented Nov 14, 2021

passive healthcheck not working after configuration changing #4077

passive healthcheck not working after configuration changing #4077

Comments

Ben0625 commented Apr 19, 2021 • edited Loading

Issue description

Environment

Minimal test code / Steps to reproduce the issue

What's the actual result?

What's the expected result?

Firstsawyou commented Apr 19, 2021

Ben0625 commented Apr 19, 2021 • edited Loading

Firstsawyou commented Apr 19, 2021

tokers commented Apr 19, 2021

Ben0625 commented Apr 20, 2021 • edited Loading

tzssangglass commented Apr 29, 2021

chzhuo commented Nov 10, 2021

shuaijinchao commented Nov 11, 2021

shuaijinchao commented Nov 11, 2021 • edited Loading

chzhuo commented Nov 12, 2021

chzhuo commented Nov 12, 2021

shuaijinchao commented Nov 14, 2021

shuaijinchao commented Nov 14, 2021

Ben0625 commented Apr 19, 2021 •

edited

Loading

Ben0625 commented Apr 19, 2021 •

edited

Loading

Ben0625 commented Apr 20, 2021 •

edited

Loading

shuaijinchao commented Nov 11, 2021 •

edited

Loading