parallelvirtualcluster
/
pvc
1
0
Fork 0
Code Issues 11 Pull Requests Projects Releases 109 Wiki Activity

Add messages to migration aborts 

Results in some information duplication, but ensures logging of the
reason a migration was aborted separate from the error(s) this may
generate.
This commit is contained in:
Joshua Boniface 2020-10-21 02:35:45 -04:00
parent d96a23276b
commit 611e47f338
1 changed files with 8 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
node-daemon/pvcnoded/VMInstance.py
View File

@ -355,7 +355,7 @@ class VMInstance(object):
aborted = False aborted = False
def abort_migrate(): def abort_migrate(reason):
zkhandler.writedata(self.zk_conn, { zkhandler.writedata(self.zk_conn, {
'/domains/{}/state'.format(self.domuuid): 'start', '/domains/{}/state'.format(self.domuuid): 'start',
'/domains/{}/node'.format(self.domuuid): self.this_node.name, '/domains/{}/node'.format(self.domuuid): self.this_node.name,
@ -363,6 +363,7 @@ class VMInstance(object):
}) })
migrate_lock_node.release() migrate_lock_node.release()
migrate_lock_state.release() migrate_lock_state.release()
self.logger.out('Aborted migration: {}'.format(reason), state='i', prefix='Domain {}'.format(self.domuuid))
# Acquire exclusive lock on the domain node key # Acquire exclusive lock on the domain node key
migrate_lock_node = zkhandler.exclusivelock(self.zk_conn, '/domains/{}/node'.format(self.domuuid)) migrate_lock_node = zkhandler.exclusivelock(self.zk_conn, '/domains/{}/node'.format(self.domuuid))
@ -374,7 +375,7 @@ class VMInstance(object):
# Don't try to migrate a node to itself, set back to start # Don't try to migrate a node to itself, set back to start
if self.node == self.lastnode or self.node == self.this_node.name: if self.node == self.lastnode or self.node == self.this_node.name:
abort_migrate() abort_migrate('Target node matches the current active node during initial check')
return return
# Synchronize nodes A (I am reader) # Synchronize nodes A (I am reader)
@ -389,7 +390,7 @@ class VMInstance(object):
time.sleep(0.1) time.sleep(0.1)
ticks += 1 ticks += 1
if ticks > 300: if ticks > 300:
self.logger.out('Timed out waiting 30s for peer, aborting migration', state='e', prefix='Domain {}'.format(self.domuuid)) self.logger.out('Timed out waiting 30s for peer', state='e', prefix='Domain {}'.format(self.domuuid))
aborted = True aborted = True
break break
self.logger.out('Releasing read lock for synchronization phase A', state='i', prefix='Domain {}'.format(self.domuuid)) self.logger.out('Releasing read lock for synchronization phase A', state='i', prefix='Domain {}'.format(self.domuuid))
@ -397,7 +398,7 @@ class VMInstance(object):
self.logger.out('Released read lock for synchronization phase A', state='o', prefix='Domain {}'.format(self.domuuid)) self.logger.out('Released read lock for synchronization phase A', state='o', prefix='Domain {}'.format(self.domuuid))
if aborted: if aborted:
abort_migrate() abort_migrate('Timed out waiting for peer')
return return
# Synchronize nodes B (I am writer) # Synchronize nodes B (I am writer)
@ -452,7 +453,7 @@ class VMInstance(object):
# Do a final verification # Do a final verification
if self.node == self.lastnode or self.node == self.this_node.name: if self.node == self.lastnode or self.node == self.this_node.name:
abort_migrate() abort_migrate('Target node matches the current active node during final check')
return return
# A live migrate is attemped 3 times in succession # A live migrate is attemped 3 times in succession
@ -468,7 +469,7 @@ class VMInstance(object):
if not migrate_live_result: if not migrate_live_result:
if force_live: if force_live:
self.logger.out('Could not live migrate VM; live migration enforced, aborting', state='e', prefix='Domain {}'.format(self.domuuid)) self.logger.out('Could not live migrate VM while live migration enforced', state='e', prefix='Domain {}'.format(self.domuuid))
aborted = True aborted = True
else: else:
do_migrate_shutdown = True do_migrate_shutdown = True
@ -478,7 +479,7 @@ class VMInstance(object):
self.logger.out('Released write lock for synchronization phase B', state='o') self.logger.out('Released write lock for synchronization phase B', state='o')
if aborted: if aborted:
abort_migrate() abort_migrate('Live migration failed and is required')
return return
# Synchronize nodes C (I am writer) # Synchronize nodes C (I am writer)