From 9f1dd284a57bc1b3d0e5d0ec7211e2431f26e054 Mon Sep 17 00:00:00 2001
From: Thomas E Lackey <telackey@bozemanpass.com>
Date: Wed, 21 Feb 2024 20:12:52 +0000
Subject: [PATCH] Better error logging for registry deployments. (#754)

We were missing errors related to registration, this should fix that.

Reviewed-on: https://git.vdb.to/cerc-io/stack-orchestrator/pulls/754
Co-authored-by: Thomas E Lackey <telackey@bozemanpass.com>
Co-committed-by: Thomas E Lackey <telackey@bozemanpass.com>
---
 .../webapp/deploy_webapp_from_registry.py     | 29 +++++++++++--------
 stack_orchestrator/deploy/webapp/util.py      | 12 ++++++--
 2 files changed, 27 insertions(+), 14 deletions(-)

diff --git a/stack_orchestrator/deploy/webapp/deploy_webapp_from_registry.py b/stack_orchestrator/deploy/webapp/deploy_webapp_from_registry.py
index 299eeba9..19497749 100644
--- a/stack_orchestrator/deploy/webapp/deploy_webapp_from_registry.py
+++ b/stack_orchestrator/deploy/webapp/deploy_webapp_from_registry.py
@@ -33,6 +33,7 @@ from stack_orchestrator.deploy.webapp.util import (LaconicRegistryClient,
 
 
 def process_app_deployment_request(
+    run_id,
     ctx,
     laconic: LaconicRegistryClient,
     app_deployment_request,
@@ -42,18 +43,8 @@ def process_app_deployment_request(
     deployment_parent_dir,
     kube_config,
     image_registry,
-    log_parent_dir
+    log_file=None
 ):
-    run_id = f"{app_deployment_request.id}-{str(time.time()).split('.')[0]}-{str(uuid.uuid4()).split('-')[0]}"
-    log_file = None
-    if log_parent_dir:
-        log_dir = os.path.join(log_parent_dir, app_deployment_request.id)
-        if not os.path.exists(log_dir):
-            os.mkdir(log_dir)
-        log_file_path = os.path.join(log_dir, f"{run_id}.log")
-        print(f"Directing build logs to: {log_file_path}")
-        log_file = open(log_file_path, "wt")
-
     # 1. look up application
     app = laconic.get_record(app_deployment_request.attributes.application, require=True)
 
@@ -291,8 +282,18 @@ def command(ctx, kube_config, laconic_config, image_registry, deployment_parent_
         for r in requests_to_execute:
             dump_known_requests(state_file, [r], "DEPLOYING")
             status = "ERROR"
+            run_log_file = None
             try:
+                run_id = f"{r.id}-{str(time.time()).split('.')[0]}-{str(uuid.uuid4()).split('-')[0]}"
+                if log_dir:
+                    run_log_dir = os.path.join(log_dir, r.id)
+                    if not os.path.exists(run_log_dir):
+                        os.mkdir(run_log_dir)
+                    run_log_file_path = os.path.join(run_log_dir, f"{run_id}.log")
+                    print(f"Directing deployment logs to: {run_log_file_path}")
+                    run_log_file = open(run_log_file_path, "wt")
                 process_app_deployment_request(
+                    run_id,
                     ctx,
                     laconic,
                     r,
@@ -302,8 +303,12 @@ def command(ctx, kube_config, laconic_config, image_registry, deployment_parent_
                     os.path.abspath(deployment_parent_dir),
                     kube_config,
                     image_registry,
-                    log_dir
+                    run_log_file
                 )
                 status = "DEPLOYED"
+            except Exception as e:
+                print("ERROR: " + str(e), file=run_log_file)
             finally:
                 dump_known_requests(state_file, [r], status)
+                if run_log_file:
+                    run_log_file.close()
diff --git a/stack_orchestrator/deploy/webapp/util.py b/stack_orchestrator/deploy/webapp/util.py
index 80b477f9..e6fd9ffe 100644
--- a/stack_orchestrator/deploy/webapp/util.py
+++ b/stack_orchestrator/deploy/webapp/util.py
@@ -227,8 +227,16 @@ def build_container_image(app_record, tag, extra_build_args=[], log_file=None):
             git_env = dict(os.environ.copy())
             # Never prompt
             git_env["GIT_TERMINAL_PROMPT"] = "0"
-            subprocess.check_call(["git", "clone", repo, clone_dir], env=git_env, stdout=log_file, stderr=log_file)
-            subprocess.check_call(["git", "checkout", ref], cwd=clone_dir, env=git_env, stdout=log_file, stderr=log_file)
+            try:
+                subprocess.check_call(["git", "clone", repo, clone_dir], env=git_env, stdout=log_file, stderr=log_file)
+            except Exception as e:
+                print(f"git clone failed.  Is the repository {repo} private?", file=log_file)
+                raise e
+            try:
+                subprocess.check_call(["git", "checkout", ref], cwd=clone_dir, env=git_env, stdout=log_file, stderr=log_file)
+            except Exception as e:
+                print(f"git checkout failed.  Does ref {ref} exist?", file=log_file)
+                raise e
         else:
             result = subprocess.run(["git", "clone", "--depth", "1", repo, clone_dir], stdout=log_file, stderr=log_file)
             result.check_returncode()