# HG changeset patch
# User emellor@xxxxxxxxxxxxxxxxxxxxxx
# Node ID a63af74b1b6ee771db316a2a52ab0eaa4326da7f
# Parent f68f6f711efc1774b96eee63839eaf0fc6ccd3dd
Change the way that reboot-timeouts are handled. Rather than refreshing the
list of domains each time XendDomainInfo.refresh is called, and spawning
threads for each shutdown still in progress, we watch the control/shutdown
node and spawn just the one thread immediately. When the daemon starts (say
after a crash) all domains are checked for entries in the store indicating a
reboot in progress.
This avoids a large cost on the store and in threads when the list of domains
is refreshed. This happens every time the server is queried, so the cost can
be quite high. It's especially a high cost when using xm shutdown -a -w --
this operation should perform a lot better now.
Signed-off-by: Ewan Mellor <ewan@xxxxxxxxxxxxx>
diff -r f68f6f711efc -r a63af74b1b6e tools/python/xen/xend/XendDomainInfo.py
--- a/tools/python/xen/xend/XendDomainInfo.py Wed Mar 8 22:57:40 2006
+++ b/tools/python/xen/xend/XendDomainInfo.py Wed Mar 8 22:57:46 2006
@@ -13,7 +13,7 @@
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
#============================================================================
# Copyright (C) 2004, 2005 Mike Wray <mike.wray@xxxxxx>
-# Copyright (C) 2005 XenSource Ltd
+# Copyright (C) 2005, 2006 XenSource Ltd
#============================================================================
"""Representation of a single domain.
@@ -82,7 +82,7 @@
STATE_DOM_OK = 1
STATE_DOM_SHUTDOWN = 2
-SHUTDOWN_TIMEOUT = 30
+SHUTDOWN_TIMEOUT = 30.0
ZOMBIE_PREFIX = 'Zombie-'
@@ -182,7 +182,7 @@
vm.initDomain()
vm.storeVmDetails()
vm.storeDomDetails()
- vm.registerWatch()
+ vm.registerWatches()
vm.refreshShutdown()
return vm
except:
@@ -238,7 +238,7 @@
vm.storeVmDetails()
vm.storeDomDetails()
- vm.registerWatch()
+ vm.registerWatches()
vm.refreshShutdown(xeninfo)
return vm
@@ -440,7 +440,10 @@
self.console_mfn = None
self.vmWatch = None
-
+ self.shutdownWatch = None
+
+ self.shutdownStartTime = None
+
self.state = STATE_DOM_OK
self.state_updated = threading.Condition()
self.refresh_shutdown_lock = threading.Condition()
@@ -644,7 +647,7 @@
self.introduceDomain()
self.storeDomDetails()
- self.registerWatch()
+ self.registerWatches()
self.refreshShutdown()
log.debug("XendDomainInfo.completeRestore done")
@@ -707,13 +710,15 @@
## public:
- def registerWatch(self):
- """Register a watch on this VM's entries in the store, so that
- when they are changed externally, we keep up to date. This should
- only be called by {@link #create}, {@link #recreate}, or {@link
- #restore}, once the domain's details have been written, but before the
- new instance is returned."""
+ def registerWatches(self):
+ """Register a watch on this VM's entries in the store, and the
+ domain's control/shutdown node, so that when they are changed
+ externally, we keep up to date. This should only be called by {@link
+ #create}, {@link #recreate}, or {@link #restore}, once the domain's
+ details have been written, but before the new instance is returned."""
self.vmWatch = xswatch(self.vmpath, self.storeChanged)
+ self.shutdownWatch = xswatch(self.dompath + '/control/shutdown',
+ self.handleShutdownWatch)
def getDomid(self):
@@ -840,33 +845,49 @@
# Domain is alive. If we are shutting it down, then check
# the timeout on that, and destroy it if necessary.
- sst = self.readDom('xend/shutdown_start_time')
- if sst:
- sst = float(sst)
- timeout = SHUTDOWN_TIMEOUT - time.time() + sst
+ if self.shutdownStartTime:
+ timeout = (SHUTDOWN_TIMEOUT - time.time() +
+ self.shutdownStartTime)
if timeout < 0:
log.info(
"Domain shutdown timeout expired: name=%s id=%s",
self.info['name'], self.domid)
self.destroy()
- else:
- log.debug(
- "Scheduling refreshShutdown on domain %d in %ds.",
- self.domid, timeout)
- threading.Timer(timeout, self.refreshShutdown).start()
finally:
self.refresh_shutdown_lock.release()
if restart_reason:
self.maybeRestart(restart_reason)
+
+
+ def handleShutdownWatch(self, _):
+ log.debug('XendDomainInfo.handleShutdownWatch')
+
+ reason = self.readDom('control/shutdown')
+
+ if reason and reason != 'suspend':
+ sst = self.readDom('xend/shutdown_start_time')
+ now = time.time()
+ if sst:
+ self.shutdownStartTime = float(sst)
+ timeout = float(sst) + SHUTDOWN_TIMEOUT - now
+ else:
+ self.shutdownStartTime = now
+ self.storeDom('xend/shutdown_start_time', now)
+ timeout = SHUTDOWN_TIMEOUT
+
+ log.trace(
+ "Scheduling refreshShutdown on domain %d in %ds.",
+ self.domid, timeout)
+ threading.Timer(timeout, self.refreshShutdown).start()
+
+ return 1
def shutdown(self, reason):
if not reason in shutdown_reasons.values():
raise XendError('Invalid reason: %s' % reason)
self.storeDom("control/shutdown", reason)
- if reason != 'suspend':
- self.storeDom('xend/shutdown_start_time', time.time())
## private:
@@ -1213,6 +1234,8 @@
"""Cleanup domain resources; release devices. Idempotent. Nothrow
guarantee."""
+ self.unwatchShutdown()
+
self.release_devices()
if self.image:
@@ -1262,6 +1285,20 @@
self.vmWatch = None
except:
log.exception("Unwatching VM path failed.")
+
+
+ def unwatchShutdown(self):
+ """Remove the watch on the domain's control/shutdown node, if any.
+ Idempotent. Nothrow guarantee."""
+
+ try:
+ try:
+ if self.shutdownWatch:
+ self.shutdownWatch.unwatch()
+ finally:
+ self.shutdownWatch = None
+ except:
+ log.exception("Unwatching control/shutdown failed.")
## public:
_______________________________________________
Xen-changelog mailing list
Xen-changelog@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-changelog
|