def handle(self, *args, **kwargs):
"""Reindexing work.
Creates a Tasktree that creates new indexes
over the old ones so the search feature
works while the indexation occurs
"""
if not django_settings.MARKETPLACE:
raise CommandError('This command affects both the marketplace and '
'AMO ES storage. But the command can only be '
'run from the Marketplace.')
force = kwargs.get('force', False)
if database_flagged() and not force:
raise CommandError('Indexation already occuring - use --force to '
'bypass')
prefix = kwargs.get('prefix', '')
log('Starting the reindexation')
if kwargs.get('wipe', False):
confirm = raw_input("Are you sure you want to wipe all data from "
"ES ? (yes/no): ")
while confirm not in ('yes', 'no'):
confirm = raw_input('Please enter either "yes" or "no": ')
if confirm == 'yes':
unflag_database()
requests.delete(url('/'))
else:
raise CommandError("Aborted.")
elif force:
unflag_database()
# Get list current aliases at /_aliases.
all_aliases = requests.get(url('/_aliases')).json()
# building the list of indexes
indexes = set([prefix + index for index in
_ALIASES.values()])
actions = []
def add_action(*elmt):
if elmt in actions:
return
actions.append(elmt)
all_aliases = all_aliases.items()
# creating a task tree
log('Building the task tree')
tree = TaskTree()
last_action = None
to_remove = []
# for each index, we create a new time-stamped index
for alias in indexes:
is_stats = 'stats' in alias
old_index = None
for aliased_index, alias_ in all_aliases:
if alias in alias_['aliases'].keys():
# mark the index to be removed later
old_index = aliased_index
to_remove.append(aliased_index)
# mark the alias to be removed as well
add_action('remove', aliased_index, alias)
# create a new index, using the alias name with a timestamp
new_index = timestamp_index(alias)
# if old_index is None that could mean it's a full index
# In that case we want to continue index in it
future_alias = url('/%s' % alias)
if requests.head(future_alias).status_code == 200:
old_index = alias
# flag the database
step1 = tree.add_task(flag_database, args=[new_index, old_index,
alias])
step2 = step1.add_task(create_mapping, args=[new_index, alias])
step3 = step2.add_task(create_index, args=[new_index, is_stats])
last_action = step3
# adding new index to the alias
add_action('add', new_index, alias)
# Alias the new index and remove the old aliases, if any.
renaming_step = last_action.add_task(run_aliases_actions,
args=[actions])
# unflag the database - there's no need to duplicate the
# indexing anymore
delete = renaming_step.add_task(unflag_database)
# Delete the old indexes, if any
delete.add_task(delete_indexes, args=[to_remove])
# let's do it
log('Running all indexation tasks')
os.environ['FORCE_INDEXING'] = '1'
try:
tree.apply_async()
time.sleep(10) # give celeryd some time to flag the DB
while database_flagged():
sys.stdout.write('.')
sys.stdout.flush()
time.sleep(5)
finally:
del os.environ['FORCE_INDEXING']
sys.stdout.write('\n')
# let's return the /_aliases values
aliases = call_es('_aliases').json()
aliases = json.dumps(aliases, sort_keys=True, indent=4)
return _SUMMARY % (len(indexes), aliases)
def handle(self, *args, **kwargs):
"""Reindexing work.
Creates a Tasktree that creates new indexes
over the old ones so the search feature
works while the indexation occurs.
"""
force = kwargs.get('force', False)
if is_reindexing_amo() and not force:
raise CommandError('Indexation already occuring - use --force to '
'bypass')
log('Starting the reindexation', stdout=self.stdout)
modules = get_modules(with_stats=kwargs.get('with_stats', False))
if kwargs.get('wipe', False):
skip_confirmation = kwargs.get('noinput', False)
confirm = ''
if not skip_confirmation:
confirm = raw_input('Are you sure you want to wipe all AMO '
'Elasticsearch indexes? (yes/no): ')
while confirm not in ('yes', 'no'):
confirm = raw_input('Please enter either "yes" or "no": ')
if (confirm == 'yes' or skip_confirmation):
unflag_database(stdout=self.stdout)
for index in set(modules.keys()):
ES.indices.delete(index, ignore=404)
else:
raise CommandError("Aborted.")
elif force:
unflag_database(stdout=self.stdout)
alias_actions = []
def add_alias_action(action, index, alias):
action = {action: {'index': index, 'alias': alias}}
if action in alias_actions:
return
alias_actions.append(action)
# Creating a task tree.
log('Building the task tree', stdout=self.stdout)
tree = TaskTree()
last_action = None
to_remove = []
# For each alias, we create a new time-stamped index.
for alias, module in modules.items():
old_index = None
olds = ES.indices.get_aliases(alias, ignore=404)
for old_index in olds:
# Mark the index to be removed later.
to_remove.append(old_index)
# Mark the alias to be removed from that index.
add_alias_action('remove', old_index, alias)
# Create a new index, using the alias name with a timestamp.
new_index = timestamp_index(alias)
# If old_index is None that could mean it's a full index.
# In that case we want to continue index in it.
if ES.indices.exists(alias):
old_index = alias
# Flag the database.
step1 = tree.add_task(flag_database,
args=[new_index, old_index, alias])
step2 = step1.add_task(create_new_index, args=[alias, new_index])
step3 = step2.add_task(index_data, args=[alias, new_index])
last_action = step3
# Adding new index to the alias.
add_alias_action('add', new_index, alias)
# Alias the new index and remove the old aliases, if any.
renaming_step = last_action.add_task(update_aliases,
args=[alias_actions])
# Unflag the database - there's no need to duplicate the
# indexing anymore.
delete = renaming_step.add_task(unflag_database)
# Delete the old indexes, if any.
if to_remove:
delete.add_task(delete_indexes, args=[to_remove])
# Let's do it.
log('Running all indexation tasks', stdout=self.stdout)
os.environ['FORCE_INDEXING'] = '1'
# This is a bit convoluted, and more complicated than simply providing
# the soft and hard time limits on the @task decorator. But we're not
# using the @task decorator here, but a decorator from celery_tasktree.
if not getattr(settings, 'CELERY_ALWAYS_EAGER', False):
control.time_limit(
'olympia.lib.es.management.commands.reindex.index_data',
soft=time_limits['soft'],
hard=time_limits['hard'])
try:
tree.apply_async()
if not getattr(settings, 'CELERY_ALWAYS_EAGER', False):
time.sleep(10) # give celeryd some time to flag the DB
while is_reindexing_amo():
sys.stdout.write('.')
sys.stdout.flush()
time.sleep(5)
finally:
del os.environ['FORCE_INDEXING']
sys.stdout.write('\n')
# Let's return the /_aliases values.
aliases = ES.indices.get_aliases()
aliases = json.dumps(aliases, sort_keys=True, indent=4)
summary = _SUMMARY % (len(modules), aliases)
log(summary, stdout=self.stdout)
def handle(self, *args, **kwargs):
"""Set up reindexing tasks.
Creates a Tasktree that creates a new indexes and indexes all objects,
then points the alias to this new index when finished.
"""
if not settings.MARKETPLACE:
raise CommandError('This command affects only marketplace and '
'should be run under Marketplace settings.')
force = kwargs.get('force', False)
prefix = kwargs.get('prefix', '')
if database_flagged() and not force:
raise CommandError('Indexation already occuring - use --force to '
'bypass')
elif force:
unflag_database()
# The list of indexes that is currently aliased by `ALIAS`.
aliases = ES.aliases(ALIAS).keys()
old_index = aliases[0] if aliases else None
# Create a new index, using the index name with a timestamp.
new_index = timestamp_index(prefix + ALIAS)
# See how the index is currently configured.
if old_index:
try:
s = (ES.get_settings(old_index).get(old_index, {})
.get('settings', {}))
except pyelasticsearch.exceptions.ElasticHttpNotFoundError:
s = {}
else:
s = {}
num_replicas = s.get('number_of_replicas',
settings.ES_DEFAULT_NUM_REPLICAS)
num_shards = s.get('number_of_shards', settings.ES_DEFAULT_NUM_SHARDS)
# Start our chain of events to re-index.
tree = TaskTree()
# Flag the database.
step1 = tree.add_task(
flag_database, args=[new_index, old_index, ALIAS])
# Create the index and mapping.
#
# Note: We set num_replicas=0 here to decrease load while re-indexing.
# In a later step we increase it which results in a more efficient bulk
# copy in Elasticsearch.
# For ES < 0.90 we manually enable compression.
step2 = step1.add_task(
create_index, args=[new_index, ALIAS,
{'number_of_replicas': 0,
'number_of_shards': num_shards,
'store.compress.tv': True,
'store.compress.stored': True,
'refresh_interval': '-1'}])
# Index all the things!
step3 = step2.add_task(run_indexing, args=[new_index])
# After indexing we optimize the index, adjust settings, and point the
# alias to the new index.
step4 = step3.add_task(
update_alias, args=[new_index, old_index, ALIAS,
{'number_of_replicas': num_replicas,
'refresh_interval': '5s'}])
# Unflag the database.
step5 = step4.add_task(unflag_database)
# Delete the old index, if any.
if old_index:
step5 = step5.add_task(delete_index, args=[old_index])
step5.add_task(output_summary)
self.stdout.write('\nNew index and indexing tasks all queued up.\n')
os.environ['FORCE_INDEXING'] = '1'
try:
tree.apply_async()
finally:
del os.environ['FORCE_INDEXING']
def handle(self, *args, **kwargs):
"""Reindexing work.
Creates a Tasktree that creates new indexes
over the old ones so the search feature
works while the indexation occurs.
"""
force = kwargs.get("force", False)
if is_reindexing_amo() and not force:
raise CommandError("Indexation already occuring - use --force to " "bypass")
log("Starting the reindexation", stdout=self.stdout)
modules = ["addons"]
if kwargs.get("with_stats", False):
modules.append("stats")
if kwargs.get("wipe", False):
skip_confirmation = kwargs.get("noinput", False)
confirm = ""
if not skip_confirmation:
confirm = raw_input("Are you sure you want to wipe all AMO " "Elasticsearch indexes? (yes/no): ")
while confirm not in ("yes", "no"):
confirm = raw_input('Please enter either "yes" or "no": ')
if confirm == "yes" or skip_confirmation:
unflag_database(stdout=self.stdout)
for index in set(MODULES[m].get_alias() for m in modules):
ES.indices.delete(index, ignore=404)
else:
raise CommandError("Aborted.")
elif force:
unflag_database(stdout=self.stdout)
alias_actions = []
def add_alias_action(action, index, alias):
action = {action: {"index": index, "alias": alias}}
if action in alias_actions:
return
alias_actions.append(action)
# Creating a task tree.
log("Building the task tree", stdout=self.stdout)
tree = TaskTree()
last_action = None
to_remove = []
# For each index, we create a new time-stamped index.
for module in modules:
old_index = None
alias = MODULES[module].get_alias()
olds = ES.indices.get_aliases(alias, ignore=404)
for old_index in olds:
# Mark the index to be removed later.
to_remove.append(old_index)
# Mark the alias to be removed from that index.
add_alias_action("remove", old_index, alias)
# Create a new index, using the alias name with a timestamp.
new_index = timestamp_index(alias)
# If old_index is None that could mean it's a full index.
# In that case we want to continue index in it.
if ES.indices.exists(alias):
old_index = alias
# Flag the database.
step1 = tree.add_task(flag_database, args=[new_index, old_index, alias])
step2 = step1.add_task(create_new_index, args=[module, new_index])
step3 = step2.add_task(index_data, args=[module, new_index])
last_action = step3
# Adding new index to the alias.
add_alias_action("add", new_index, alias)
# Alias the new index and remove the old aliases, if any.
renaming_step = last_action.add_task(update_aliases, args=[alias_actions])
# Unflag the database - there's no need to duplicate the
# indexing anymore.
delete = renaming_step.add_task(unflag_database)
# Delete the old indexes, if any.
if to_remove:
delete.add_task(delete_indexes, args=[to_remove])
# Let's do it.
log("Running all indexation tasks", stdout=self.stdout)
os.environ["FORCE_INDEXING"] = "1"
# This is a bit convoluted, and more complicated than simply providing
# the soft and hard time limits on the @task decorator. But we're not
# using the @task decorator here, but a decorator from celery_tasktree.
if not getattr(settings, "CELERY_ALWAYS_EAGER", False):
control.time_limit(
"olympia.lib.es.management.commands.reindex.index_data",
soft=time_limits["soft"],
hard=time_limits["hard"],
)
try:
tree.apply_async()
if not getattr(settings, "CELERY_ALWAYS_EAGER", False):
time.sleep(10) # give celeryd some time to flag the DB
while is_reindexing_amo():
sys.stdout.write(".")
sys.stdout.flush()
time.sleep(5)
finally:
del os.environ["FORCE_INDEXING"]
sys.stdout.write("\n")
# Let's return the /_aliases values.
aliases = ES.indices.get_aliases()
aliases = json.dumps(aliases, sort_keys=True, indent=4)
summary = _SUMMARY % (len(modules), aliases)
log(summary, stdout=self.stdout)
def handle(self, *args, **kwargs):
"""Reindexing work.
Creates a Tasktree that creates new indexes
over the old ones so the search feature
works while the indexation occurs
"""
if not django_settings.MARKETPLACE:
raise CommandError('This command affects both the marketplace and '
'AMO ES storage. But the command can only be '
'run from the Marketplace.')
force = kwargs.get('force', False)
if database_flagged() and not force:
raise CommandError('Indexation already occuring - use --force to '
'bypass')
prefix = kwargs.get('prefix', '')
log('Starting the reindexation')
if kwargs.get('wipe', False):
confirm = raw_input("Are you sure you want to wipe all data from "
"ES ? (yes/no): ")
while confirm not in ('yes', 'no'):
confirm = raw_input('Please enter either "yes" or "no": ')
if confirm == 'yes':
unflag_database()
requests.delete(url('/'))
else:
raise CommandError("Aborted.")
elif force:
unflag_database()
# Get list current aliases at /_aliases.
all_aliases = requests.get(url('/_aliases')).json
# building the list of indexes
indexes = set(
[prefix + index for index in django_settings.ES_INDEXES.values()])
actions = []
def add_action(*elmt):
if elmt in actions:
return
actions.append(elmt)
all_aliases = all_aliases.items()
# creating a task tree
log('Building the task tree')
tree = TaskTree()
last_action = None
to_remove = []
# for each index, we create a new time-stamped index
for alias in indexes:
is_stats = 'stats' in alias
old_index = None
for aliased_index, alias_ in all_aliases:
if alias in alias_['aliases'].keys():
# mark the index to be removed later
old_index = aliased_index
to_remove.append(aliased_index)
# mark the alias to be removed as well
add_action('remove', aliased_index, alias)
# create a new index, using the alias name with a timestamp
new_index = timestamp_index(alias)
# if old_index is None that could mean it's a full index
# In that case we want to continue index in it
future_alias = url('/%s' % alias)
if requests.head(future_alias).status_code == 200:
old_index = alias
# flag the database
step1 = tree.add_task(flag_database,
args=[new_index, old_index, alias])
step2 = step1.add_task(create_mapping, args=[new_index, alias])
step3 = step2.add_task(create_index, args=[new_index, is_stats])
last_action = step3
# adding new index to the alias
add_action('add', new_index, alias)
# Alias the new index and remove the old aliases, if any.
renaming_step = last_action.add_task(run_aliases_actions,
args=[actions])
# unflag the database - there's no need to duplicate the
# indexing anymore
delete = renaming_step.add_task(unflag_database)
# Delete the old indexes, if any
delete.add_task(delete_indexes, args=[to_remove])
# let's do it
log('Running all indexation tasks')
os.environ['FORCE_INDEXING'] = '1'
try:
tree.apply_async()
time.sleep(10) # give celeryd some time to flag the DB
while database_flagged():
sys.stdout.write('.')
sys.stdout.flush()
time.sleep(5)
finally:
del os.environ['FORCE_INDEXING']
sys.stdout.write('\n')
# let's return the /_aliases values
aliases = call_es('_aliases').json
aliases = json.dumps(aliases, sort_keys=True, indent=4)
return _SUMMARY % (len(indexes), aliases)
def handle(self, *args, **kwargs):
"""Reindexing work.
Creates a Tasktree that creates new indexes
over the old ones so the search feature
works while the indexation occurs.
"""
force = kwargs.get('force', False)
if is_reindexing_amo() and not force:
raise CommandError('Indexation already occuring - use --force to '
'bypass')
log('Starting the reindexation', stdout=self.stdout)
modules = get_modules(with_stats=kwargs.get('with_stats', False))
if kwargs.get('wipe', False):
skip_confirmation = kwargs.get('noinput', False)
confirm = ''
if not skip_confirmation:
confirm = raw_input('Are you sure you want to wipe all AMO '
'Elasticsearch indexes? (yes/no): ')
while confirm not in ('yes', 'no'):
confirm = raw_input('Please enter either "yes" or "no": ')
if (confirm == 'yes' or skip_confirmation):
unflag_database(stdout=self.stdout)
for index in set(modules.keys()):
ES.indices.delete(index, ignore=404)
else:
raise CommandError("Aborted.")
elif force:
unflag_database(stdout=self.stdout)
alias_actions = []
def add_alias_action(action, index, alias):
action = {action: {'index': index, 'alias': alias}}
if action in alias_actions:
return
alias_actions.append(action)
# Creating a task tree.
log('Building the task tree', stdout=self.stdout)
tree = TaskTree()
last_action = None
to_remove = []
# For each alias, we create a new time-stamped index.
for alias, module in modules.items():
old_index = None
olds = ES.indices.get_aliases(alias, ignore=404)
for old_index in olds:
# Mark the index to be removed later.
to_remove.append(old_index)
# Mark the alias to be removed from that index.
add_alias_action('remove', old_index, alias)
# Create a new index, using the alias name with a timestamp.
new_index = timestamp_index(alias)
# If old_index is None that could mean it's a full index.
# In that case we want to continue index in it.
if ES.indices.exists(alias):
old_index = alias
# Flag the database.
step1 = tree.add_task(flag_database,
args=[new_index, old_index, alias])
step2 = step1.add_task(create_new_index,
args=[alias, new_index])
step3 = step2.add_task(index_data,
args=[alias, new_index])
last_action = step3
# Adding new index to the alias.
add_alias_action('add', new_index, alias)
# Alias the new index and remove the old aliases, if any.
renaming_step = last_action.add_task(update_aliases,
args=[alias_actions])
# Unflag the database - there's no need to duplicate the
# indexing anymore.
delete = renaming_step.add_task(unflag_database)
# Delete the old indexes, if any.
if to_remove:
delete.add_task(delete_indexes, args=[to_remove])
# Let's do it.
log('Running all indexation tasks', stdout=self.stdout)
os.environ['FORCE_INDEXING'] = '1'
# This is a bit convoluted, and more complicated than simply providing
# the soft and hard time limits on the @task decorator. But we're not
# using the @task decorator here, but a decorator from celery_tasktree.
if not getattr(settings, 'CELERY_ALWAYS_EAGER', False):
control.time_limit(
'olympia.lib.es.management.commands.reindex.index_data',
soft=time_limits['soft'],
hard=time_limits['hard'])
try:
tree.apply_async()
if not getattr(settings, 'CELERY_ALWAYS_EAGER', False):
time.sleep(10) # give celeryd some time to flag the DB
while is_reindexing_amo():
sys.stdout.write('.')
sys.stdout.flush()
time.sleep(5)
finally:
del os.environ['FORCE_INDEXING']
sys.stdout.write('\n')
# Let's return the /_aliases values.
aliases = ES.indices.get_aliases()
aliases = json.dumps(aliases, sort_keys=True, indent=4)
summary = _SUMMARY % (len(modules), aliases)
log(summary, stdout=self.stdout)