From 6612bb4e3e55cb2608a0988f2bc0e219b759ca64 Mon Sep 17 00:00:00 2001
From: Yaron Haviv <yaronh@iguaz.io>
Date: Thu, 1 Jul 2021 01:27:25 +0300
Subject: [PATCH] [API] Fix bugs when using feature vectors and in real time
 ingestion (#1073)

---
 mlrun/execution.py         | 15 +++++++++++----
 mlrun/feature_store/api.py |  2 +-
 mlrun/runtimes/base.py     |  4 +++-
 mlrun/runtimes/function.py | 18 ++++++++++++++++--
 4 files changed, 31 insertions(+), 8 deletions(-)

diff --git a/mlrun/execution.py b/mlrun/execution.py
index a12dc5a4d4c..96955206109 100644
--- a/mlrun/execution.py
+++ b/mlrun/execution.py
@@ -233,7 +233,14 @@ def get_meta(self):
 
     @classmethod
     def from_dict(
-        cls, attrs: dict, rundb="", autocommit=False, tmp="", host=None, log_stream=None
+        cls,
+        attrs: dict,
+        rundb="",
+        autocommit=False,
+        tmp="",
+        host=None,
+        log_stream=None,
+        is_api=False,
     ):
         """create execution context from dict"""
 
@@ -269,15 +276,15 @@ def from_dict(
 
         self._init_dbs(rundb)
 
-        if spec:
-            # init data related objects (require DB & Secrets to be set first)
+        if spec and not is_api:
+            # init data related objects (require DB & Secrets to be set first), skip when running in the api service
             self._data_stores.from_dict(spec)
             if inputs and isinstance(inputs, dict):
                 for k, v in inputs.items():
                     if v:
                         self._set_input(k, v)
 
-        if host:
+        if host and not is_api:
             self.set_label("host", host)
 
         start = get_in(attrs, "status.start_time")
diff --git a/mlrun/feature_store/api.py b/mlrun/feature_store/api.py
index dcaf2aa5411..85cac9b66df 100644
--- a/mlrun/feature_store/api.py
+++ b/mlrun/feature_store/api.py
@@ -433,7 +433,7 @@ def deploy_ingestion_service(
         featureset, source, targets, run_config.parameters
     )
 
-    name = name or f"{featureset.metadata.name}_ingest"
+    name = name or f"{featureset.metadata.name}-ingest"
     if not run_config.function:
         function_ref = featureset.spec.function.copy()
         if function_ref.is_empty():
diff --git a/mlrun/runtimes/base.py b/mlrun/runtimes/base.py
index 68cab7e659c..bfbcd8e5bd5 100644
--- a/mlrun/runtimes/base.py
+++ b/mlrun/runtimes/base.py
@@ -435,7 +435,9 @@ def run(
                 "warning!, Api url not set, " "trying to exec remote runtime locally"
             )
 
-        execution = MLClientCtx.from_dict(runspec.to_dict(), db, autocommit=False)
+        execution = MLClientCtx.from_dict(
+            runspec.to_dict(), db, autocommit=False, is_api=self._is_api_server
+        )
         self._pre_run(runspec, execution)  # hook for runtime specific prep
 
         # create task generator (for child runs) from spec
diff --git a/mlrun/runtimes/function.py b/mlrun/runtimes/function.py
index 8a8b57ac5cd..d8513d49b39 100644
--- a/mlrun/runtimes/function.py
+++ b/mlrun/runtimes/function.py
@@ -371,6 +371,8 @@ def add_vault_config_to_spec(self):
     def deploy(
         self, dashboard="", project="", tag="", verbose=False,
     ):
+        # todo: verify that the function name is normalized
+
         verbose = verbose or self.verbose
         if verbose:
             self.set_env("MLRUN_LOG_LEVEL", "DEBUG")
@@ -738,6 +740,11 @@ def deploy_nuclio_function(function: RemoteRuntime, dashboard="", watch=False):
         spec.set_config("spec.resources", function.spec.resources)
     if function.spec.no_cache:
         spec.set_config("spec.build.noCache", True)
+    if function.spec.build.functionSourceCode:
+        spec.set_config(
+            "spec.build.functionSourceCode", function.spec.build.functionSourceCode
+        )
+
     if function.spec.replicas:
         spec.set_config("spec.minReplicas", function.spec.replicas)
         spec.set_config("spec.maxReplicas", function.spec.replicas)
@@ -746,9 +753,16 @@ def deploy_nuclio_function(function: RemoteRuntime, dashboard="", watch=False):
         spec.set_config("spec.maxReplicas", function.spec.max_replicas)
 
     dashboard = dashboard or mlconf.nuclio_dashboard_url
-    if function.spec.base_spec:
+    if function.spec.base_spec or function.spec.build.functionSourceCode:
+        config = function.spec.base_spec
+        if not config:
+            # if base_spec was not set (when not using code_to_function) and we have base64 code
+            # we create the base spec with essential attributes
+            config = nuclio.config.new_config()
+            update_in(config, "spec.handler", handler or "main:handler")
+
         config = nuclio.config.extend_config(
-            function.spec.base_spec, spec, tag, function.spec.build.code_origin
+            config, spec, tag, function.spec.build.code_origin
         )
         update_in(config, "metadata.name", function.metadata.name)
         update_in(config, "spec.volumes", function.spec.generate_nuclio_volumes())