"automatic merge of lp:~bladernr/checkbox/1612220-fix-disk_smart/ by tarmac [r=sylvain-pineau][bug=1612220][author=bladernr]"

1 files changed, 34 insertions, 11 deletions

diff --git a/bin/disk_smart b/bin/disk_smart
index a753da5d..804b11fd 100755
--- a/bin/disk_smart
+++ b/bin/disk_smart
@@ -188,10 +188,12 @@ def count_raid_disks(disk):
 
 
 def initiate_smart_test(disk, raid_element, raid_type, type='short'):
+ # Note, '-t force' ensures we abort any existing smart test in progress
+ # and start a clean run.
 if raid_type == 'none':
- ctl_command = 'smartctl -t {} {}'.format(type, disk)
+ ctl_command = 'smartctl -t {} -t force {}'.format(type, disk)
 else:
- ctl_command = ('smartctl -t {} {} -d {},{}'.
+ ctl_command = ('smartctl -t {} -t force {} -d {},{}'.
 format(type, disk, raid_type, raid_element))
 logging.debug('Beginning test with {}'.format(ctl_command))
 
@@ -217,6 +219,9 @@ def get_smart_entries(disk, raid_element, raid_type, type='selftest'):
 universal_newlines=True)
 returncode = 0
 except CalledProcessError as err:
+ logging.error("Error encountered checking SMART Log")
+ logging.error("\tsmartctl returned code: {}".format(err.returncode))
+ logging.error("\tSee 'man smartctl' for info on return code meanings")
 stdout = err.output
 returncode = err.returncode
 
@@ -264,11 +269,14 @@ def in_progress(current_entries):
 :returns:
 True if an "in-progress" message is found, False otherwise
 """
+ # LP:1612220 Only check first log entry for status to avoid false triggers
+ # on older interrupted tests that may still show an "in progress" status.
 statuses = [entry for entry in current_entries
- if isinstance(entry, dict)
- and 'status' in entry
- and (entry['status'] == 'Self-test routine in progress'
- or "Self test in progress" in entry['status'])]
+ if isinstance(entry,
+ dict) and 'status' in entry and
+ entry['number'] == 1 and (
+ entry['status'] == 'Self-test routine in progress' or
+ "Self test in progress" in entry['status'])]
 if statuses:
 for entry in statuses:
 logging.debug('%s %s %s %s' % (entry['number'],
@@ -304,7 +312,7 @@ def poll_for_status(args, disk, raid_element, raid_type, previous_entries):
 """
 # Priming read... this is here in case our test is finished or fails
 # immediate after it beginsAccording to.
- logging.debug('Polling selftest.log for status')
+ logging.debug('Polling SMART selftest log for status')
 keep_going = True
 
 while keep_going:
@@ -315,6 +323,8 @@ def poll_for_status(args, disk, raid_element, raid_type, previous_entries):
 raid_type)
 if current_entries != previous_entries:
 if not in_progress(current_entries):
+ logging.debug("Current log entries differ from starting log"
+ " entries. Stopping polling.")
 keep_going = False
 
 if args.timeout is not None:
@@ -358,6 +368,12 @@ def run_smart_test(args, disk, raid_element, raid_type):
 if len(previous_entries) > 20:
 # Abort the previous instance
 # so that polling can identify the difference
+
+ # The proper way to kill the test is using -X
+ # kill_smart_test(disk, raid_element, raid_type)
+ # Then re-initiate the test
+ logging.debug("Log is 20+ entries long. Restarting test to add an"
+ " abort message to make the log diff easier")
 initiate_smart_test(disk, raid_element, raid_type)
 previous_entries, returncode = get_smart_entries(disk, raid_element,
 raid_type)
@@ -376,8 +392,15 @@ def run_smart_test(args, disk, raid_element, raid_type):
 "for some reason. Run 'sudo smartctl -l selftest "
 "{} -d {},{}' to see the SMART log".
 format(disk, raid_type, raid_element))
- logging.debug("Last smartctl return code: %d", returncode)
- logging.debug("Last smartctl run status: %s", status)
+ logging.error("\tLast smartctl return code: %d", returncode)
+ logging.error("\tLast smartctl run status: %s", status)
+ logging.debug("\tMost Recent SMART LOG Entry:")
+ for log_entry in log:
+ if log_entry['number'] == 1:
+ logging.debug("\t# {}\t{}\t{}\t{}\t{}\t{}".format(
+ log_entry['number'], log_entry['description'],
+ log_entry['status'], log_entry['remaining'],
+ log_entry['lifetime'], log_entry['lba']))
 return False
 else:
 if raid_type == 'none':
@@ -438,8 +461,8 @@ def main():
 success = True
 for raid_element in range(0, num_disks):
 if enable_smart(disk, raid_element, raid_type):
- success = (run_smart_test(args, disk, raid_element, raid_type)
- and success)
+ success = (run_smart_test(args, disk, raid_element,
+ raid_type) and success)
 else:
 success = False
 if success is False: