THREESCALE-6552 - RFE #1255

samugi · 2021-01-26T21:26:59Z

Adds the ability to select a list of upstrams (backends) to apply maintenance mode to + key/value liquid enabled operations.
Updates a variable in the liquid context when the Upstream Policy changes the Upstream.

samugi · 2021-01-26T21:27:50Z

gateway/src/apicast/policy/upstream/upstream.lua

@@ -58,6 +58,10 @@ function _M:rewrite(context)
  end
 end

+function _M:access(context)
+  context.route_upstream = context[self] or context.route_upstream


is this ok?

so basically you're exposing the upstream URL defined in the the proxy_config during the access phase here right? Do we do this anywhere else? One quick test to check this doesn't break anything is to run the entire test suite locally.

yeah we are doing it here in the routing policy. I thought it would be nice to have a default variable in the context where, should a policy modify the upstream, it would store the new value, so I used the same variable here. This way we would know where to look for an updated upstream, regardless of which policy changed it. The reason I ask is I am not sure this was intended to be used this way and maybe I should use a dedicated variable for the Upstream policy instead.
According to the tests it doesn't break anything at the moment :P

palmieric · 2021-01-27T21:08:24Z

gateway/src/apicast/policy/maintenance_mode/maintenance_mode.lua

+  local path2 = uri2.path or ""
+  path2 = string.gsub(path2, "/$", "")
+
+  return scheme1 == scheme2 and uri1.host == uri2.host and port1 == port2 and path1 == path2 


Why are you checking the path too?

@palmieric fixed it, I was supposed to "match" the paths instead of using ==. Thanks!

kevprice83 · 2021-01-28T13:17:52Z

gateway/src/apicast/policy/upstream/upstream.lua

@@ -58,6 +58,10 @@ function _M:rewrite(context)
  end
 end

+function _M:access(context)
+  context.route_upstream = context[self] or context.route_upstream


so basically you're exposing the upstream URL defined in the the proxy_config during the access phase here right? Do we do this anywhere else? One quick test to check this doesn't break anything is to run the entire test suite locally.

kevprice83 · 2021-01-28T13:19:21Z

spec/policy/maintenance_mode/maintenance_mode_spec.lua

    before_each(function()
+      ctx = { service = { api_backend = 'http://backend2.example.org:80' } }
+      ctx2 = {


does this mean we have 2 different instances of the context object during runtime? that could be problematic no as from anywhere we only ever reference it as context or ctx

my intention here was to use a different context (ctx2) just for some of the tests. I guess I can also achieve this by modifying the same ctx object from within the test where I need a different value.

fixed this as per our discussion, I also added a couple more tests.

eloycoto

Now only nitpicks! I'm happy to see this PR :-)

eloycoto · 2021-02-09T21:08:44Z

gateway/src/apicast/policy/maintenance_mode/maintenance_mode.lua

@@ -19,20 +22,54 @@ function _M.new(configuration)
  policy.message_content_type = default_message_content_type

  if configuration then
+    policy.maintenance_upstreams = configuration.upstreams


This is no longer needed no?

yup, forgot it :)

eloycoto · 2021-02-09T21:10:31Z

gateway/src/apicast/policy/maintenance_mode/maintenance_mode.lua

  ngx.header['Content-Type'] = self.message_content_type
  ngx.status = self.status_code
  ngx.say(self.message)

  return ngx.exit(ngx.status)
 end

+function _M:access(context)
+  -- Before the condition was not set, so the maintenance mode when condition is


In this past, this policy does not have condition. So by default, if condition is not set it set the maintenance mode.

eloycoto · 2021-02-09T21:11:04Z

gateway/src/apicast/policy/maintenance_mode/maintenance_mode.lua

+  end
+
+  context.upstream = context.route_upstream or context:get_upstream() or {}
+  context.upstream = context.upstream.uri or context.upstream


This is incorrect, no?

local upstream = context.route_upstream or context:get_upstream() or {} context.upstream = upstream.uri or upstream

This is a bit better

eloycoto · 2021-02-09T21:11:56Z

spec/policy/maintenance_mode/maintenance_mode_spec.lua

+    context('when Maintenance Mode is configured for a specific upstream', function()  
+      local maintenance_policy = MaintenancePolicy.new({
+        condition = {
+          operations={{op="==", left="{{ upstream.scheme }}://{{ upstream.host }}{{ upstream.port }}{{ upstream.path }}", left_type="liquid", right="https://"..test_host.."443"..test_path, right_type="plain"}},


I would indent this a bit, in all testsuite TBH.

eloycoto · 2021-02-09T21:12:50Z

spec/policy/maintenance_mode/maintenance_mode_spec.lua

+      end)
+    end)
+
+    context('when Maintenance Mode is configured for a specific upstream (different host)', function()  


This can be all in the it expression TBH, and the same for all others

eloycoto · 2021-02-09T21:17:10Z

t/apicast-policy-maintenance-mode.t

 --- response_body
 { "msg": "Be back soon" }
 --- response_headers
 Content-Type: application/json
 --- error_code: 503
 --- no_error_log
 [error]
+
+
+=== TEST 5: Maintenance mode is applied with routing policy + matching upstream condition


I'm missing here the following test:

-> Works with api-backend, I think that it's not longer used by system, but just in case.
-> A OR condition, where the first is not matching, but matches the second one. It's covered by unit, but I prefer to have here too.

Regards

eloycoto · 2021-02-12T15:45:25Z

Add a Changelog and it'll be merged :-)

samugi · 2021-02-12T19:09:18Z

Add a Changelog and it'll be merged :-)

all done!

samugi requested review from eloycoto, kevprice83 and palmieric January 26, 2021 21:26

samugi requested a review from a team as a code owner January 26, 2021 21:26

samugi commented Jan 26, 2021

View reviewed changes

palmieric reviewed Jan 27, 2021

View reviewed changes

samugi force-pushed the THREESCALE-6552 branch from 43ee5e6 to 0593a5e Compare January 28, 2021 11:22

kevprice83 reviewed Jan 28, 2021

View reviewed changes

samugi force-pushed the THREESCALE-6552 branch from 0593a5e to b20cb03 Compare January 29, 2021 21:20

samugi requested a review from kevprice83 January 29, 2021 21:32

samugi force-pushed the THREESCALE-6552 branch 2 times, most recently from 49d4245 to 0330364 Compare February 7, 2021 17:54

eloycoto suggested changes Feb 9, 2021

View reviewed changes

samugi force-pushed the THREESCALE-6552 branch from 0330364 to 5ec3282 Compare February 10, 2021 20:33

maintenance mode with liquid upstream filtering

e677830

samugi force-pushed the THREESCALE-6552 branch from 5ec3282 to e677830 Compare February 12, 2021 19:07

samugi requested a review from eloycoto February 12, 2021 19:09

eloycoto approved these changes Feb 15, 2021

View reviewed changes

eloycoto merged commit d655f4f into 3scale:master Feb 15, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

THREESCALE-6552 - RFE #1255

THREESCALE-6552 - RFE #1255

samugi commented Jan 26, 2021

samugi Jan 26, 2021

kevprice83 Jan 28, 2021

samugi Jan 29, 2021

palmieric Jan 27, 2021

samugi Jan 28, 2021

kevprice83 Jan 28, 2021

kevprice83 Jan 28, 2021

samugi Jan 28, 2021

samugi Jan 29, 2021

eloycoto left a comment

eloycoto Feb 9, 2021

samugi Feb 10, 2021

eloycoto Feb 9, 2021

eloycoto Feb 9, 2021

eloycoto Feb 10, 2021

eloycoto Feb 9, 2021

eloycoto Feb 9, 2021

eloycoto Feb 9, 2021

eloycoto commented Feb 12, 2021

samugi commented Feb 12, 2021

THREESCALE-6552 - RFE #1255

THREESCALE-6552 - RFE #1255

Conversation

samugi commented Jan 26, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

eloycoto left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

eloycoto commented Feb 12, 2021

samugi commented Feb 12, 2021