def __init__(self, creds, master, slaves, backup_lineage, thread_lock=None, do_force_reset=False, do_set_dns_slave=True, refresh_lineage=None, timestamp_override=None, progress_output=True):
threading.Thread.__init__(self)
self.entry_exit_state = None
self.server = Server(int(creds['rs_api_account_id']), creds['rs_api_user'], creds['rs_api_password'])
self.ebss = EbsSnapshot(int(creds['rs_api_account_id']), creds['rs_api_user'], creds['rs_api_password'])
self.ebsv = EbsVolume(int(creds['rs_api_account_id']), creds['rs_api_user'], creds['rs_api_password'])
self.master = master
self.slaves = slaves
self.do_force_reset = do_force_reset
self.do_set_dns_slave = do_set_dns_slave
self.backup_lineage = backup_lineage
self.refresh_lineage = refresh_lineage
self.timestamp_override = timestamp_override
self.progress_output = progress_output
if not thread_lock:
self.thread_lock = threading.Lock()
else:
self.thread_lock = thread_lock
self.blocking = False
class DbRefreshHelper(threading.Thread):
def __init__(self, creds, master, slaves, backup_lineage, thread_lock=None, do_force_reset=False, do_set_dns_slave=True, refresh_lineage=None, timestamp_override=None, progress_output=True):
threading.Thread.__init__(self)
self.entry_exit_state = None
self.server = Server(int(creds['rs_api_account_id']), creds['rs_api_user'], creds['rs_api_password'])
self.ebss = EbsSnapshot(int(creds['rs_api_account_id']), creds['rs_api_user'], creds['rs_api_password'])
self.ebsv = EbsVolume(int(creds['rs_api_account_id']), creds['rs_api_user'], creds['rs_api_password'])
self.master = master
self.slaves = slaves
self.do_force_reset = do_force_reset
self.do_set_dns_slave = do_set_dns_slave
self.backup_lineage = backup_lineage
self.refresh_lineage = refresh_lineage
self.timestamp_override = timestamp_override
self.progress_output = progress_output
if not thread_lock:
self.thread_lock = threading.Lock()
else:
self.thread_lock = thread_lock
self.blocking = False
def run(self):
self.server.logger.info("Starting to refresh the databases...")
# Get lock to synchronize threads
self.thread_lock.acquire(self.blocking)
if self.do_force_reset == True:
f_servers = []
f_servers.append(self.master)
if self.slaves:
for s in self.slaves:
f_servers.append(s)
failed_servers = self.force_reset(f_servers)
if failed_servers:
self.server.logger.error("Failed to force reset servers. Quitting as the refresh will not work properly")
self.entry_exit_state = 'failed'
return
## give it a moment....
self.server.logger.info("Sleeping for a moment...")
time.sleep(15)
mserver = self.server.get_details({'nickname': self.master})
master_instance_href = mserver[0]['current_instance_href']
if not self.restore_master():
self.entry_exit_state = 'failed'
return
## if there are no slaves, then we're done here.
if not self.slaves:
self.entry_exit_state = 'successful'
return
if not self.backup_master():
self.entry_exit_state = 'failed'
return
polling_threads = []
volumes = self.ebsv.get_details({'ec2_instance_href': master_instance_href})
ebs_nickname = self.ebsv.get_nickname(volumes)
thread_lock = threading.Lock()
db_thread = Auditing.AuditSnapshot(thread_lock, self.ebss, ebs_nickname, progress_output=self.progress_output)
db_thread.start()
polling_threads.append(db_thread)
if polling_threads:
map(lambda t: t.join(), polling_threads)
latest_timestamp = self.ebss.get_latest_timestamp(self.server, self.master)
self.ebss.logger.info("Latest Snapshot Timestamp on %s: %s" % (self.master, latest_timestamp))
time.sleep(5)
if not self.restore_slaves(latest_timestamp):
self.entry_exit_state = 'failed'
return
# Free lock to release next thread
if self.thread_lock.locked():
self.thread_lock.release()
self.entry_exit_state = 'successful'
return
def restore_master(self):
polling_threads = []
thread_lock = threading.Lock()
db_thread = None
recipe = 'db::do_primary_restore_and_set_master'
if self.timestamp_override and self.refresh_lineage:
self.server.logger.info("Restoring %s using refresh lineage %s and timestamp override: %s" %
(self.master, self.refresh_lineage, self.timestamp_override))
recipe_data = {
'db/backup/lineage': "text:%s" % self.refresh_lineage,
'db/backup/timestamp_override': "text:%s" % self.timestamp_override
}
else:
if self.refresh_lineage:
self.server.logger.info("Restoring %s using backup lineage: %s" % (self.master, self.refresh_lineage))
recipe_data = {
'db/backup/lineage': "text:%s" % self.refresh_lineage,
}
else:
self.server.logger.info("Restoring %s using timestamp override: %s" % (self.master, self.timestamp_override))
recipe_data = {
'db/backup/lineage': "text:%s" % self.refresh_lineage,
'db/backup/timestamp_override': "text:%s" % self.timestamp_override
}
if self.server.run_recipe(self.master, recipe, recipe_data):
location = self.server.get_header_location()
if location:
db_thread = Auditing.AuditEntry(thread_lock, self.server, self.master, recipe, location, 20, progress_output=self.progress_output)
db_thread.start()
polling_threads.append(db_thread)
else:
self.server.logger.error("Unable to restore %s!" % self.master)
return False
if polling_threads:
map(lambda t: t.join(), polling_threads)
if not db_thread or db_thread.entry_exit_state == 'failed':
return False
self.server.logger.info("Sleeping for a couple seconds to allow the volumes to settle...")
time.sleep(20)
return True
def backup_master(self):
polling_threads = []
thread_lock = threading.Lock()
retries = 2
current_try = 0
recipe = 'db::do_primary_backup'
recipe_data = {
'db/backup/lineage': "text:%s" % self.backup_lineage
}
final_status = None
location = None
db_thread = None
while (current_try <= retries):
current_try = current_try + 1
if self.server.run_recipe(self.master, recipe, recipe_data):
location = self.server.get_header_location()
if location:
db_thread = Auditing.AuditEntry(thread_lock, self.server, self.master, recipe, location, progress_output=self.progress_output)
db_thread.start()
polling_threads.append(db_thread)
else:
## if we fail to run the recipe, we're done.
self.server.logger.error("Failed to run %s on %s, quitting!" % (recipe, self.master))
return False
if polling_threads:
map(lambda t: t.join(), polling_threads)
if db_thread and db_thread.entry_exit_state:
final_status = db_thread.entry_exit_state
if final_status == 'completed':
break
if current_try <= retries:
self.server.logger.warn("Trying again in 5 seconds...")
time.sleep(5)
if not final_status or final_status == 'failed':
return False
self.server.logger.info("Sleeping for a couple seconds to allow the volumes to settle...")
time.sleep(20)
return True
def restore_slaves(self, master_timestamp):
polling_threads = []
db_threads = []
thread_lock = threading.Lock()
db_thread = None
recipe = 'db::do_primary_init_slave'
recipe_data = {
'db/backup/lineage': "text:%s" % self.backup_lineage,
'db/backup/timestamp_override': "text:%s" % master_timestamp
}
recipes_after = ['db::request_master_allow']
if self.do_set_dns_slave:
recipes_after.append('db::do_set_dns_slave')
failed_servers = []
failed_servers_recipes = []
## let's do the init slave first....
for slave in self.slaves:
location = None
if self.server.run_recipe(slave, recipe, recipe_data):
location = self.server.get_header_location()
if location:
db_thread = Auditing.AuditEntry(thread_lock, self.server, slave, recipe, location, 20, progress_output=self.progress_output)
db_thread.start()
_thread_data = {slave: db_thread}
db_threads.append(_thread_data)
else:
failed_servers.append(slave)
if db_threads:
for d in db_threads:
for slave, thread in d.iteritems():
thread.join()
for d in db_threads:
for slave, thread in d.iteritems():
if thread.entry_exit_state == 'failed':
self.server.logger.error("%s failed for %s" % (recipe, slave))
## should not be in there already, but just making sure...
if not slave in failed_servers:
failed_servers.append(slave)
## and now we'll do the rest of the recipe's
for slave in self.slaves:
if not slave in failed_servers:
for r in recipes_after:
location = None
if self.server.run_recipe(slave, r):
location = self.server.get_header_location()
if location:
db_thread = Auditing.AuditEntry(thread_lock, self.server, slave, r, location, progress_output=self.progress_output)
db_thread.start()
polling_threads.append(db_thread)
else:
_fail_data = {slave: {'recipe': r}}
failed_servers_recipes.append(_fail_data)
if polling_threads:
map(lambda t: t.join(), polling_threads)
## if those recipes failed, then log a warning
if failed_servers_recipes:
for f in failed_servers_recipes:
for _server, data in f.iteritems():
self.server.logger.warn("Recipe \"%s\" failed on server \"%s\"" % (data['recipe'], _server))
## only return false if we were unable to refresh a slave. This does constitute a failure.
if failed_servers:
self.server.logger.error("We failed to refresh: %s" % ', '.join(failed_servers))
return False
self.server.logger.info("Slaves data has been restored from the master.")
return True
def force_reset(self, reset_servers, retry=1):
thread_lock = threading.Lock()
retry_limit = 2
polling_threads = []
successful_servers = []
failed_servers = []
recipe = 'db::do_force_reset'
data = {'db/force_safety': 'text:off'}
self.server.logger.info("Force Resetting: %s" % ', '.join(reset_servers))
for s in reset_servers:
if self.server.run_recipe(s, recipe, data):
location = self.server.get_header_location()
if location:
db_thread = Auditing.AuditEntry(thread_lock, self.server, s, recipe, location, 20, progress_output=self.progress_output)
db_thread.start()
thread_data = {
'nickname': s,
'thread': db_thread
}
polling_threads.append(thread_data)
else:
failed_servers.append(s)
if polling_threads:
for t in polling_threads:
t['thread'].join()
if t['thread'].entry_exit_state == 'failed':
if t['nickname'] not in failed_servers:
failed_servers.append(t['nickname'])
else:
if t['nickname'] not in successful_servers:
successful_servers.append(t['nickname'])
## fixes the false failure when recursion happens because failed_servers persists
if successful_servers:
for s in successful_servers:
if s in failed_servers:
failed_servers.remove(s)
if len(failed_servers) > 0 and retry <= retry_limit:
retry += 1
self.server.logger.warn("Server(s): %s - failed to force reset, trying again..." % ', '.join(failed_servers))
time.sleep(5)
self.force_reset(failed_servers, retry)
if successful_servers:
self.server.logger.info("Successfully Forced Reset: %s" % ', '.join(successful_servers))
if failed_servers:
self.server.logger.warn("Failed to force reset: %s" % ', '.join(failed_servers))
if retry > retry_limit:
self.server.logger.warn("We've exceeded the retry limit for force resetting these servers!")
return failed_servers
