[CI] Raise [keystone_authtoken]http_request_max_retries

This is in the hope to avoid random "Unable to validate token" in
AIO upgrade jobs which restart the Keystone container while
other services are trying to use it (most notably Placement but
also Nova and Neutron).

Change-Id: Ie6a95c3ca84df82ca8463cf76054e125cb1ffff1

tests/check-logs.sh

@@ -28,15 +28,10 @@ function filter_out_expected_critical {
     # job.
 
     case $1 in
-    */placement-api.log)
-        # Sometimes we see this during upgrade when keystone is down.
-        grep -v "Failed to fetch token data from identity server"
-        ;;
     */neutron-server.log)
         # Sometimes we see this during shutdown (upgrade).
         # See: https://bugs.launchpad.net/neutron/+bug/1863579
         grep -v "WSREP has not yet prepared node for application use"
-        grep -v "Failed to fetch token data from identity server"
         ;;
     *)
         # We have to provide some pass-through consumer to avoid:

tests/run.yml

@@ -162,6 +162,10 @@
         # globals.yml
         - src: "tests/templates/globals-default.j2"
           dest: /etc/kolla/globals.yml
+        # global.conf
+        - src: "tests/templates/global.conf.j2"
+          dest: /etc/kolla/config/global.conf
+          when: "{{ openstack_core_enabled }}"
         # nova-compute.conf
         - src: "tests/templates/nova-compute-overrides.j2"
           dest: /etc/kolla/config/nova/nova-compute.conf

tests/templates/global.conf.j2

@@ -0,0 +1,7 @@
+[keystone_authtoken]
+# NOTE(yoctozepto): This is to avoid CRITICAL messages in logs
+# when services try to contact Keystone when it is down due
+# to upgrade running, seen especially in AIO jobs, where there
+# is no backup to respond, most often with Placement, but also
+# with Neutron and Nova.
+http_request_max_retries = 9