From 1d2721cd898abf6fde89f4fed60a85e40cf592c4 Mon Sep 17 00:00:00 2001
From: Luca Miccini <lmiccini@redhat.com>
Date: Wed, 23 Oct 2024 17:46:41 +0200
Subject: [PATCH] Add delay parameter

We've seen issues in the past with nfs servers that would not release locks fast enough for evacuation to complete (see https://bugzilla.redhat.com/show_bug.cgi?id=1755760 for example).
This commit adds a DELAY parameter so we can wait for a certain amount of time before starting the evacuation.

We also take the chance to fix a typo (hardcoded value vs variable).
---
 templates/instanceha/bin/instanceha.py  | 6 +++++-
 templates/instanceha/config/config.yaml | 1 +
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/templates/instanceha/bin/instanceha.py b/templates/instanceha/bin/instanceha.py
index ddd23010..9995c51a 100755
--- a/templates/instanceha/bin/instanceha.py
+++ b/templates/instanceha/bin/instanceha.py
@@ -68,6 +68,7 @@ def start_health_check_server():
 TAGGED_IMAGES = config["TAGGED_IMAGES"] if 'TAGGED_IMAGES' in config else "true"
 TAGGED_FLAVORS = config["TAGGED_FLAVORS"] if 'TAGGED_FLAVORS' in config else "true"
 DELTA = int(config["DELTA"]) if 'DELTA' in config else 30
+DELAY = int(config["DELAY"]) if 'DELAY' in config else 0
 POLL = int(config["POLL"]) if 'POLL' in config else 45
 THRESHOLD = int(config["THRESHOLD"]) if 'THRESHOLD' in config else 50
 WORKERS = int(config["WORKERS"]) if 'WORKERS' in config else 4
@@ -182,6 +183,9 @@ def _host_evacuate(connection, host):
         logging.info("Nothing to evacuate")
         return True
 
+    # sleep for DELAY, this could be useful if nfs is in use and locks need to be released
+    time.sleep(DELAY)
+
     # if SMART_EVACUATION is 'True' (string) use a ThreadPoolExecutor to poll the evacuation status
     # otherwise use the old "fire and forget" approach
     if 'true' in SMART_EVACUATION.lower():
@@ -865,7 +869,7 @@ def main():
                         with concurrent.futures.ThreadPoolExecutor() as executor:
                             results = list(executor.map(lambda service: process_service(service, reserved_hosts, True), to_resume))
                         if not all(results):
-                            logging.warning('Some services failed to evacuate. Retrying in 30 seconds.')
+                            logging.warning('Some services failed to evacuate. Retrying in %s seconds.' % POLL)
 
                     else:
                         logging.info('InstanceHa DISABLE is true, not evacuating')
diff --git a/templates/instanceha/config/config.yaml b/templates/instanceha/config/config.yaml
index a8c5fecc..7367d23c 100644
--- a/templates/instanceha/config/config.yaml
+++ b/templates/instanceha/config/config.yaml
@@ -4,6 +4,7 @@ config:
   TAGGED_FLAVORS: "true"
   SMART_EVACUATION: "false"
   DELTA: "30"
+  DELAY: "0"
   POLL: "45"
   THRESHOLD: "50"
   WORKERS: "4"