Merge pull request #385 from Cray-HPE/CASMCMS-9162-release-2.10

CASMCMS-9162: Add cfs_read_timeout option

CHANGELOG.md

@@ -5,7 +5,12 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
-## [Unreleased]
+## [2.10.28] - 2024-10-16
+### Added
+#### BOS option
+- cfs_read_timeout: Allow the amount of time BOS waits for a response from CFS before
+  timing out to be configurable
+
 ### Fixed
 - Fix per-bootset CFS setting
 

api/openapi.yaml.in

@@ -1603,6 +1603,13 @@ components:
         Options for the Boot Orchestration Service.
       type: object
       properties:
+        cfs_read_timeout:
+          type: integer
+          description: |
+            The amount of time (in seconds) to wait for a response before timing out a request to CFS
+          example: 20
+          minimum: 10
+          maximum: 86400
         cleanup_completed_session_ttl:
           type: string
           description: |

src/bos/operators/utils/clients/bos/options.py

@@ -71,6 +71,10 @@ def get_option(self, key, value_type, default):
         else:
             raise KeyError('Option {} not found and no default exists'.format(key))
 
+    @property
+    def cfs_read_timeout(self):
+        return self.get_option('cfs_read_timeout', int, 10)
+
     @property
     def logging_level(self):
         return self.get_option('logging_level', str, 'INFO')

src/bos/operators/utils/clients/cfs.py

@@ -23,7 +23,8 @@
 #
 from collections import defaultdict
 import logging
-from requests.exceptions import HTTPError, ConnectionError
+from bos.operators.utils.clients.bos.options import options
+from requests.exceptions import HTTPError
 
 from bos.common.utils import compact_response_text, exc_type_msg, requests_retry_session, PROTOCOL
 
@@ -45,7 +46,7 @@ def get_components(session=None, **params):
     Returns the list of CFS components
     """
     if not session:
-        session = requests_retry_session()
+        session = requests_retry_session(read_timeout=options.cfs_read_timeout)  # pylint: disable=redundant-keyword-arg
     component_list = []
     while params is not None:
         LOGGER.debug("GET %s with params=%s", COMPONENTS_ENDPOINT, params)
@@ -71,7 +72,7 @@ def patch_components(data, session=None):
         LOGGER.warning("patch_components called without data; returning without action.")
         return
     if not session:
-        session = requests_retry_session()
+        session = requests_retry_session(read_timeout=options.cfs_read_timeout)  # pylint: disable=redundant-keyword-arg
     LOGGER.debug("PATCH %s with body=%s", COMPONENTS_ENDPOINT, data)
     response = session.patch(COMPONENTS_ENDPOINT, json=data)
     LOGGER.debug("Response status code=%d, reason=%s, body=%s", response.status_code,
@@ -88,7 +89,7 @@ def get_components_from_id_list(id_list):
         LOGGER.warning("get_components_from_id_list called without IDs; returning without action.")
         return []
     LOGGER.debug("get_components_from_id_list called with %d IDs", len(id_list))
-    session = requests_retry_session()
+    session = requests_retry_session(read_timeout=options.cfs_read_timeout)  # pylint: disable=redundant-keyword-arg
     component_list = []
     while id_list:
         next_batch = id_list[:GET_BATCH_SIZE]
@@ -106,7 +107,7 @@ def patch_desired_config(node_ids, desired_config, enabled=False, tags=None, cle
         return
     LOGGER.debug("patch_desired_config called on %d IDs with desired_config=%s enabled=%s tags=%s"
                  " clear_state=%s", len(node_ids), desired_config, enabled, tags, clear_state)
-    session = requests_retry_session()
+    session = requests_retry_session(read_timeout=options.cfs_read_timeout)  # pylint: disable=redundant-keyword-arg
     node_patch = {
         'enabled': enabled,
         'desired_config': desired_config,

src/bos/server/controllers/v2/options.py

@@ -37,6 +37,7 @@
 # options simpler
 OPTIONS_KEY = 'options'
 DEFAULTS = {
+    'cfs_read_timeout': 10,
     'cleanup_completed_session_ttl': "7d",
     'clear_stage': False,
     'component_actual_state_ttl': "4h",