parallelvirtualcluster
/
pvc
1
0
Fork 0
Code Issues 11 Pull Requests Projects Releases 111 Wiki Activity

Use transitional takeover states for migration 

Use a pair of transitional states, "takeover" and "relinquish", when
transitioning between primary and secondary coordinator states. This
provides a clsuter-wide record that the nodes are still working during
their synchronous transition states, and should allow clients to
determine when the node(s) have fully switched over. Also add an
additional 2 seconds of wait at the end of the transition jobs to ensure
everything has had a chance to start before proceeding.

References #72
This commit is contained in:
Joshua Boniface 2020-02-19 13:18:38 -05:00
parent 8678dedfea
commit d2a5fe59c0
2 changed files with 27 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
node-daemon/pvcnoded/Daemon.py
View File

@ -575,14 +575,17 @@ def cleanup():
pass
# Force into secondary network state if needed
if zkhandler.readdata(zk_conn, '/nodes/{}/routerstate'.format(myhostname)) == 'primary':
is_primary = True
zkhandler.writedata(zk_conn, {
'/nodes/{}/routerstate'.format(myhostname): 'secondary',
'/primary_node': 'none'
})
logger.out('Waiting 5 seconds for primary migration', state='s')
time.sleep(5)
try:
if this_node.router_state == 'primary':
is_primary = True
zkhandler.writedata(zk_conn, {
'/primary_node': 'none'
})
logger.out('Waiting for primary migration', state='s')
while this_node.router_state != 'secondary':
time.sleep(1)
except:
pass
# Set stop state in Zookeeper
zkhandler.writedata(zk_conn, { '/nodes/{}/daemonstate'.format(myhostname): 'stop' })
@ -825,9 +828,10 @@ def update_primary(new_primary, stat, event=''):
logger.out('Contending for primary coordinator state', state='i')
zkhandler.writedata(zk_conn, {'/primary_node': myhostname})
elif new_primary == myhostname:
zkhandler.writedata(zk_conn, {'/nodes/{}/routerstate'.format(myhostname): 'primary'})
zkhandler.writedata(zk_conn, {'/nodes/{}/routerstate'.format(myhostname): 'takeover'})
else:
zkhandler.writedata(zk_conn, {'/nodes/{}/routerstate'.format(myhostname): 'secondary'})
if this_node.router_state != 'secondary':
zkhandler.writedata(zk_conn, {'/nodes/{}/routerstate'.format(myhostname): 'relinquish'})
else:
zkhandler.writedata(zk_conn, {'/nodes/{}/routerstate'.format(myhostname): 'client'})

17
node-daemon/pvcnoded/NodeInstance.py
View File

@ -117,16 +117,19 @@ class NodeInstance(object):
if data != self.router_state:
self.router_state = data
if self.config['enable_networking']:
if self.router_state == 'primary':
if self.router_state == 'takeover':
self.logger.out('Setting node {} to primary state'.format(self.name), state='i')
transition_thread = threading.Thread(target=self.become_primary, args=(), kwargs={})
transition_thread.start()
else:
if self.router_state == 'relinquish':
# Skip becoming secondary unless already running
if self.daemon_state == 'run' or self.daemon_state == 'shutdown':
self.logger.out('Setting node {} to secondary state'.format(self.name), state='i')
transition_thread = threading.Thread(target=self.become_secondary, args=(), kwargs={})
transition_thread.start()
else:
# We did nothing, so just become secondary state
zkhandler.writedata(self.zk_conn, {'/nodes/{}/routerstate'.format(self.name): 'secondary'})
@self.zk_conn.DataWatch('/nodes/{}/domainstate'.format(self.name))
def watch_node_domainstate(data, stat, event=''):
@ -428,8 +431,8 @@ class NodeInstance(object):
self.logger.out('Setting Patroni leader to this node', state='i')
tick = 1
patroni_failed = True
# As long as we're primary, keep trying to set the Patroni leader to us
while self.router_state == 'primary':
# As long as we're in takeover, keep trying to set the Patroni leader to us
while self.router_state == 'takeover':
# Switch Patroni leader to the local instance
retcode, stdout, stderr = common.run_os_command(
"""
@ -489,7 +492,10 @@ class NodeInstance(object):
lock.release()
self.logger.out('Released write lock for synchronization G', state='o')
# Wait 2 seconds for everything to stabilize before we declare all-done
time.sleep(2)
primary_lock.release()
zkhandler.writedata(self.zk_conn, {'/nodes/{}/routerstate'.format(self.name): 'primary'})
self.logger.out('Node {} transitioned to primary state'.format(self.name), state='o')
def become_secondary(self):
@ -611,6 +617,9 @@ class NodeInstance(object):
lock.release()
self.logger.out('Released read lock for synchronization G', state='o')
# Wait 2 seconds for everything to stabilize before we declare all-done
time.sleep(2)
zkhandler.writedata(self.zk_conn, {'/nodes/{}/routerstate'.format(self.name): 'secondary'})
self.logger.out('Node {} transitioned to secondary state'.format(self.name), state='o')
# Flush all VMs on the host