def test_registerLauncher(self):
launcher = zk.Launcher()
launcher.id = "launcher-000-001"
self.zk.registerLauncher(launcher)
launchers = self.zk.getRegisteredLaunchers()
self.assertEqual(1, len(launchers))
self.assertEqual(launcher.id, launchers[0].id)
def run(self):
self.running = True
while self.running:
# Don't do work if we've lost communication with the ZK cluster
did_suspend = False
while self.zk and (self.zk.suspended or self.zk.lost):
did_suspend = True
self.log.info("ZooKeeper suspended. Waiting")
time.sleep(SUSPEND_WAIT_TIME)
if did_suspend:
self.log.info("ZooKeeper available. Resuming")
# Make sure we're always registered with ZK
launcher = zk.Launcher()
launcher.id = self.launcher_id
for prov_cfg in self.nodepool.config.providers.values():
launcher.supported_labels.update(prov_cfg.getSupportedLabels())
launcher.provider_name = self.provider_name
self.zk.registerLauncher(launcher)
self.updateProviderLimits(
self.nodepool.config.providers.get(self.provider_name))
try:
if not self.paused_handler:
while not self._assignHandlers():
# _assignHandlers can take quite some time on a busy
# system so sprinkle _removeCompletedHandlers in
# between such that we have a chance to fulfill
# requests that already have all nodes.
self._removeCompletedHandlers()
else:
# If we are paused, one request handler could not
# satisfy its assigned request, so give it
# another shot. Unpause ourselves if it completed.
self.paused_handler.run()
if not self.paused_handler.paused:
self.paused_handler = None
self._removeCompletedHandlers()
except Exception:
self.log.exception("Error in PoolWorker:")
self.stop_event.wait(self.watermark_sleep)
# Cleanup on exit
if self.paused_handler:
self.paused_handler.unlockNodeSet(clear_allocation=True)
