enh(bam): sometimes kpi creation may break BA's consistency

REFS: MON-34895

broker/bam/src/ba.cc

@@ -619,7 +619,7 @@ void ba::update_from(computable* child, io::stream* visitor) {
                       kpi_child->get_id());
   bool changed = _apply_changes(kpi_child, new_hard_impact, new_soft_impact,
                                 kpi_in_downtime);
-  SPDLOG_LOGGER_TRACE(log_v2::bam(), "BA has changed: {}", changed);
+  SPDLOG_LOGGER_TRACE(log_v2::bam(), "BA {} has changed: {}", _id, changed);
 
   // Check for inherited downtimes.
   _compute_inherited_downtime(visitor);

broker/bam/src/configuration/applier/kpi.cc

@@ -93,10 +93,10 @@ void applier::kpi::apply(bam::configuration::state::kpis const& my_kpis,
   bam::configuration::state::kpis to_create(my_kpis);
 
   // Iterate through configuration.
-  for (bam::configuration::state::kpis::iterator it(to_create.begin()),
-       end(to_create.end());
+  for (bam::configuration::state::kpis::iterator it = to_create.begin(),
+                                                 end = to_create.end();
        it != end;) {
-    std::map<uint32_t, applied>::iterator cfg_it(to_delete.find(it->first));
+    std::map<uint32_t, applied>::iterator cfg_it = to_delete.find(it->first);
     // Found = modify (or not).
     if (cfg_it != to_delete.end()) {
       // Configuration mismatch, modify object
@@ -354,6 +354,8 @@ void applier::kpi::_resolve_kpi(configuration::kpi const& cfg,
 
   my_ba->add_impact(kpi);
   kpi->add_parent(my_ba);
+  /* The propagation is forced to be sure all the ba is coherent. */
+  my_ba->notify_parents_of_change(nullptr);
 }
 
 /**

init.sh

@@ -0,0 +1,22 @@
+#!/bin/bash
+# A little script to source before compiling.
+# It just sets VCPKG_ROOT and PATH so that make/ninja won't fail.
+
+# To be sure the script is sourced.
+(return 0 2>/dev/null) && sourced=1 || sourced=0
+
+if (( sourced == 0 )) ; then
+  echo "Please execute this script with:"
+  echo ". init.sh"
+  exit 1
+fi
+
+# To be sure the script has not already been launched.
+if [[ "$PATH" =~ "vcpkg" ]] ; then
+  echo "Already initialized"
+  exit 2
+fi
+
+# The main purpose.
+export VCPKG_ROOT=$PWD/vcpkg
+export PATH=$VCPKG_ROOT:$PATH

tests/bam/bam_pb.robot

@@ -1125,6 +1125,147 @@ BA_RATIO_PERCENT_BA_4_SERVICE
     [Teardown]    Run Keywords    Ctn Stop engine    AND    Ctn Kindly Stop Broker
 
 
+BA_CHANGED
+    [Documentation]    A BA of type worst is configured with one service kpi.
+    ...                Then it is modified so that the service kpi is replaced
+    ...                by a boolean rule kpi. When cbd is reloaded, the BA is
+    ...                well updated.
+    [Tags]    MON-34895
+    Ctn Bam Init
+
+    @{svc}    Set Variable    ${{ [("host_16", "service_302")] }}
+    ${ba}    Ctn Create Ba With Services    test    worst    ${svc}
+
+    Ctn Start Broker
+    ${start}    Get Current Date
+    Ctn Start Engine
+    # Let's wait for the external command check start
+    ${content}    Create List    check_for_external_commands()
+    ${result}    Ctn Find In Log With Timeout    ${engineLog0}    ${start}    ${content}    60
+    Should Be True    ${result}    A message telling check_for_external_commands() should be available.
+
+    # Both services ${state} => The BA parent is ${state}
+    Ctn Process Service Result Hard
+    ...    host_16
+    ...    service_302
+    ...    0
+    ...    output OK for service 302
+
+    ${result}    Ctn Check Ba Status With Timeout    test    0    30
+    Ctn Dump Ba On Error    ${result}    ${ba[0]}
+    Should Be True    ${result}    The BA test is not OK as expected
+
+    Ctn Remove Service Kpi    ${ba[0]}    host_16    service_302
+    Ctn Add Boolean Kpi
+    ...    ${ba[0]}
+    ...    {host_16 service_302} {IS} {OK}
+    ...    False
+    ...    100
+
+    Ctn Reload Broker
+    Remove File    /tmp/ba.dot
+    Ctn Broker Get Ba    51001    ${ba[0]}    /tmp/ba.dot
+    Wait Until Created    /tmp/ba.dot
+    ${result}    Grep File    /tmp/ba.dot    Boolean exp
+    Should Not Be Empty    ${result}
+
+    Ctn Add Boolean Kpi
+    ...    ${ba[0]}
+    ...    {host_16 service_303} {IS} {WARNING}
+    ...    False
+    ...    100
+
+    Ctn Reload Broker
+    Remove File    /tmp/ba.dot
+    Ctn Broker Get Ba    51001    ${ba[0]}    /tmp/ba.dot
+    Wait Until Created    /tmp/ba.dot
+    ${result}    Grep File    /tmp/ba.dot    BOOL Service (16, 303)
+    Should Not Be Empty    ${result}
+    [Teardown]    Run Keywords    Ctn Stop engine    AND    Ctn Kindly Stop Broker
+
+
+BA_IMPACT_IMPACT
+    [Documentation]    A BA of type impact is defined with two BAs of type impact
+    ...                as children. The first child has an impact of 90 and the
+    ...                second one of 10. When they are impacting both, the
+    ...                parent should be critical. When they are not impacting,
+    ...                the parent should be ok.
+    [Tags]    MON-34895
+    Ctn Bam Init
+
+    ${parent_ba}    Ctn Create Ba    parent    impact    20    99
+    @{svc1}    Set Variable    ${{ [("host_16", "service_302")] }}
+    ${child1_ba}    Ctn Create Ba    child1    impact    20    99
+    Ctn Add Service Kpi    host_16    service_302    ${child1_ba[0]}    100    2    3
+    ${child2_ba}    Ctn Create Ba    child2    impact    20    99
+    Ctn Add Service Kpi    host_16    service_303    ${child2_ba[0]}    100    2    3
+
+    Ctn Add Ba Kpi    ${child1_ba[0]}    ${parent_ba[0]}    90    2    3
+    Ctn Add Ba Kpi    ${child2_ba[0]}    ${parent_ba[0]}    10    2    3
+
+    Ctn Start Broker
+    ${start}    Get Current Date
+    Ctn Start Engine
+    # Let's wait for the external command check start
+    ${content}    Create List    check_for_external_commands()
+    ${result}    Ctn Find In Log With Timeout    ${engineLog0}    ${start}    ${content}    60
+    Should Be True    ${result}    A message telling check_for_external_commands() should be available.
+
+    FOR    ${state}    ${value}    IN
+    ...    OK          0
+    ...    CRITICAL    2
+    ...    OK          0
+    ...    CRITICAL    2
+        # Both services ${state} => The BA parent is ${state}
+        Ctn Process Service Result Hard
+        ...    host_16
+        ...    service_302
+        ...    ${value}
+        ...    output ${state} for service 302
+
+        Ctn Process Service Result Hard
+        ...    host_16
+        ...    service_303
+        ...    ${value}
+        ...    output ${state} for service 302
+
+        ${result}    Ctn Check Service Status With Timeout    host_16    service_302    ${value}    60    HARD
+        Should Be True    ${result}    The service (host_16,service_302) is not ${state} as expected
+        ${result}    Ctn Check Service Status With Timeout    host_16    service_303    ${value}    60    HARD
+        Should Be True    ${result}    The service (host_16,service_303) is not ${state} as expected
+
+        ${result}    Ctn Check Ba Status With Timeout    child1    ${value}    30
+        Ctn Dump Ba On Error    ${result}    ${child1_ba[0]}
+        Should Be True    ${result}    The BA child1 is not ${state} as expected
+
+        ${result}    Ctn Check Ba Status With Timeout    child2    ${value}    30
+        Ctn Dump Ba On Error    ${result}    ${child2_ba[0]}
+        Should Be True    ${result}    The BA child2 is not ${state} as expected
+
+        ${result}    Ctn Check Ba Status With Timeout    parent    ${value}    30
+        Ctn Dump Ba On Error    ${result}    ${parent_ba[0]}
+        Should Be True    ${result}    The BA parent is not ${state} as expected
+
+        Remove Files    /tmp/parent1.dot    /tmp/parent2.dot
+        Ctn Broker Get Ba    51001    ${parent_ba[0]}    /tmp/parent1.dot
+        Wait Until Created    /tmp/parent1.dot
+
+        ${start}    Get Current Date
+        Ctn Reload Broker
+        ${content}    Create List    Inherited downtimes and BA states restored
+        ${result}    Ctn Find In Log With Timeout    ${centralLog}    ${start}    ${content}    60
+        Should Be True    ${result}    It seems that no cache has been restored into BAM.
+
+        Ctn Broker Get Ba    51001    ${parent_ba[0]}    /tmp/parent2.dot
+        Wait Until Created    /tmp/parent2.dot
+
+        ${result}    Ctn Compare Dot Files    /tmp/parent1.dot    /tmp/parent2.dot
+        Should Be True    ${result}    The BA changed during Broker reload.
+    END
+
+    [Teardown]    Run Keywords    Ctn Stop engine    AND    Ctn Kindly Stop Broker
+
+
 *** Keywords ***
 Ctn BAM Setup
     Ctn Stop Processes

tests/resources/Engine.py

@@ -1339,6 +1339,19 @@ def ctn_add_service_kpi(host: str, serv: str, id_ba: int, critical_impact: int,
         host, serv, id_ba, critical_impact, warning_impact, unknown_impact)
 
 
+def ctn_remove_service_kpi(id_ba: int, host: str, svc: str):
+    """
+    Remove a service kpi given by hostname/service description from a ba given by its id.
+
+    Args:
+        id_ba: The BA ID.
+        host: the host name.
+        svc: the service description.
+    """
+    global dbconf
+    dbconf.ctn_remove_service_kpi(id_ba, host, svc)
+
+
 def ctn_get_command_id(service: int):
     """
     Get the command ID of the service with the given ID.

tests/resources/db_conf.py

@@ -286,6 +286,19 @@ def ctn_add_service_kpi(self, host: str, serv: str, id_ba: int, critical_impact:
 
             connection.commit()
 
+    def ctn_remove_service_kpi(self, id_ba: int, host: str, svc: str):
+        connection = pymysql.connect(host=DB_HOST,
+                                     user=DB_USER,
+                                     password=DB_PASS,
+                                     database=DB_NAME_CONF,
+                                     charset='utf8mb4',
+                                     cursorclass=pymysql.cursors.DictCursor)
+
+        with connection:
+            with connection.cursor() as cursor:
+                cursor.execute(f"DELETE FROM mod_bam_kpi WHERE host_id={self.host[host]} AND service_id={self.service[svc]} AND id_ba={id_ba}")
+            connection.commit()
+
     def ctn_add_boolean_kpi(self, id_ba: int, expression: str, impact_if: bool, critical_impact: int):
         connection = pymysql.connect(host=DB_HOST,
                                      user=DB_USER,