[Notes] [Git][BuildStream/buildstream][mablanch/630-remote-execution-reconn] 6 commits: _artifactcache: Rename update_atime() to update_mtime()

[Martin Blanchard <gitlab mg gitlab com>]

Title: GitLab

Martin Blanchard pushed to branch mablanch/630-remote-execution-reconn at BuildStream / buildstream

Commits:

• 12719f0d
  by Jürg Billeter at 2018-10-22T17:05:41Z

  _artifactcache: Rename update_atime() to update_mtime()
  
  os.utime() updates both, however, list_artifacts() sorts refs by mtime,
  i.e., atime is irrelevant.
  

• a7a28d14
  by Jürg Billeter at 2018-10-22T17:05:41Z

  _artifactcache: Fix ref in update_mtime()
  
  get_artifact_fullname() is required to construct the ref. The cache key
  alone does not suffice.
  

• be29e0f5
  by Jürg Billeter at 2018-10-22T17:53:26Z

  Merge branch 'juerg/cas-mtime' into 'master'
  
  _artifactcache: Fix ref in update_mtime()
  
  See merge request BuildStream/buildstream!882

• b74aca1a
  by Jürg Billeter at 2018-10-23T09:22:19Z

  tests/testutils/artifactshare.py: Fix has_artifact() to match core code
  
  Replace path separator with '-' to match Element.normal_name.
  

• c7dda150
  by Jürg Billeter at 2018-10-23T09:48:00Z

  Merge branch 'juerg/test-artifactshare' into 'master'
  
  tests/testutils/artifactshare.py: Fix has_artifact() to match core code
  
  See merge request BuildStream/buildstream!890

• 797d3010
  by Martin Blanchard at 2018-10-23T10:01:38Z

  _sandboxremote.py: Try to reopen operation steam on failure
  
  The REAPI allows a client to reconnect to an ongoing operation stream by
  providing a WaitExecution(). If implemented on server side, BuildStream
  will try to recover from connection errors using it.
  
  https://gitlab.com/BuildStream/buildstream/issues/630
  

4 changed files:

• buildstream/_artifactcache/artifactcache.py
• buildstream/_artifactcache/cascache.py
• buildstream/sandbox/_sandboxremote.py
• tests/testutils/artifactshare.py

Changes:

buildstream/_artifactcache/artifactcache.py

---

...	...	@@ -228,7 +228,7 @@ class ArtifactCache():
228	228	self._required_elements.update(elements)
229	229
230	230	# For the cache keys which were resolved so far, we bump
231		- # the atime of them.
	231	+ # the mtime of them.
232	232	#
233	233	# This is just in case we have concurrent instances of
234	234	# BuildStream running with the same artifact cache, it will
...	...	@@ -240,7 +240,7 @@ class ArtifactCache():
240	240	for key in (strong_key, weak_key):
241	241	if key:
242	242	try:
243		- self.update_atime(key)
	243	+ self.update_mtime(element, key)
244	244	except ArtifactError:
245	245	pass
246	246
...	...	@@ -391,15 +391,16 @@ class ArtifactCache():
391	391	def preflight(self):
392	392	pass
393	393
394		- # update_atime()
	394	+ # update_mtime()
395	395	#
396		- # Update the atime of an artifact.
	396	+ # Update the mtime of an artifact.
397	397	#
398	398	# Args:
	399	+ # element (Element): The Element to update
399	400	# key (str): The key of the artifact.
400	401	#
401		- def update_atime(self, key):
402		- raise ImplError("Cache '{kind}' does not implement contains()"
	402	+ def update_mtime(self, element, key):
	403	+ raise ImplError("Cache '{kind}' does not implement update_mtime()"
403	404	.format(kind=type(self).__name__))
404	405
405	406	# initialize_remotes():

buildstream/_artifactcache/cascache.py

---

...	...	@@ -538,8 +538,9 @@ class CASCache(ArtifactCache):
538	538	except FileNotFoundError as e:
539	539	raise ArtifactError("Attempt to access unavailable artifact: {}".format(e)) from e
540	540
541		- def update_atime(self, ref):
	541	+ def update_mtime(self, element, key):
542	542	try:
	543	+ ref = self.get_artifact_fullname(element, key)
543	544	os.utime(self._refpath(ref))
544	545	except FileNotFoundError as e:
545	546	raise ArtifactError("Attempt to access unavailable artifact: {}".format(e)) from e

buildstream/sandbox/_sandboxremote.py

---

...	...	@@ -76,8 +76,7 @@ class SandboxRemote(Sandbox):
76	76	# Upload the Command message to the remote CAS server
77	77	command_digest = cascache.push_message(self._get_project(), remote_command)
78	78	if not command_digest or not cascache.verify_digest_pushed(self._get_project(), command_digest):
79		- # Command push failed
80		- return None
	79	+ raise SandboxError("Failed pushing build command to remote CAS.")
81	80
82	81	# Create and send the action.
83	82	action = remote_execution_pb2.Action(command_digest=command_digest,
...	...	@@ -88,27 +87,57 @@ class SandboxRemote(Sandbox):
88	87	# Upload the Action message to the remote CAS server
89	88	action_digest = cascache.push_message(self._get_project(), action)
90	89	if not action_digest or not cascache.verify_digest_pushed(self._get_project(), action_digest):
91		- # Action push failed
92		- return None
	90	+ raise SandboxError("Failed pushing build action to remote CAS.")
93	91
94	92	# Next, try to create a communication channel to the BuildGrid server.
95	93	channel = grpc.insecure_channel(self.server_url)
96	94	stub = remote_execution_pb2_grpc.ExecutionStub(channel)
97	95	request = remote_execution_pb2.ExecuteRequest(action_digest=action_digest,
98	96	skip_cache_lookup=False)
99		- try:
100		- operation_iterator = stub.Execute(request)
101		- except grpc.RpcError:
102		- return None
	97	+
	98	+ def __run_remote_command(stub, execute_request=None, running_operation=None):
	99	+ try:
	100	+ last_operation = None
	101	+ if execute_request is not None:
	102	+ operation_iterator = stub.Execute(execute_request)
	103	+ else:
	104	+ request = remote_execution_pb2.WaitExecutionRequest(name=running_operation.name)
	105	+ operation_iterator = stub.WaitExecution(request)
	106	+
	107	+ for operation in operation_iterator:
	108	+ if operation.done:
	109	+ return operation
	110	+ else:
	111	+ last_operation = operation
	112	+ except grpc.RpcError as e:
	113	+ status_code = e.code()
	114	+ if status_code == grpc.StatusCode.UNAVAILABLE:
	115	+ raise SandboxError("Failed contacting remote execution server at {}."
	116	+ .format(self.server_url))
	117	+
	118	+ elif status_code in (grpc.StatusCode.INVALID_ARGUMENT,
	119	+ grpc.StatusCode.FAILED_PRECONDITION,
	120	+ grpc.StatusCode.RESOURCE_EXHAUSTED,
	121	+ grpc.StatusCode.INTERNAL,
	122	+ grpc.StatusCode.DEADLINE_EXCEEDED):
	123	+ raise SandboxError("{} ({}).".format(e.details(), status_code.name))
	124	+
	125	+ elif running_operation and status_code == grpc.StatusCode.UNIMPLEMENTED:
	126	+ raise SandboxError("Failed trying to recover from connection loss: "
	127	+ "server does not support operation status polling recovery.")
	128	+
	129	+ return last_operation
103	130
104	131	operation = None
105	132	with self._get_context().timed_activity("Waiting for the remote build to complete"):
106		- # It is advantageous to check operation_iterator.code() is grpc.StatusCode.OK here,
107		- # which will check the server is actually contactable. However, calling it when the
108		- # server is available seems to cause .code() to hang forever.
109		- for operation in operation_iterator:
110		- if operation.done:
111		- break
	133	+ operation = __run_remote_command(stub, execute_request=request)
	134	+ if operation is None:
	135	+ return None
	136	+ elif operation.done:
	137	+ return operation
	138	+
	139	+ while operation is not None and not operation.done:
	140	+ operation = __run_remote_command(stub, running_operation=operation)
112	141
113	142	return operation
114	143
...	...	@@ -192,7 +221,6 @@ class SandboxRemote(Sandbox):
192	221
193	222	if operation is None:
194	223	# Failure of remote execution, usually due to an error in BuildStream
195		- # NB This error could be raised in __run_remote_command
196	224	raise SandboxError("No response returned from server")
197	225
198	226	assert not operation.HasField('error') and operation.HasField('response')

tests/testutils/artifactshare.py

---

...	...	@@ -122,9 +122,8 @@ class ArtifactShare():
122	122	# same algo for creating an artifact reference
123	123	#
124	124
125		- # Chop off the .bst suffix first
126		- assert element_name.endswith('.bst')
127		- element_name = element_name[:-4]
	125	+ # Replace path separator and chop off the .bst suffix
	126	+ element_name = os.path.splitext(element_name.replace(os.sep, '-'))[0]
128	127
129	128	valid_chars = string.digits + string.ascii_letters + '-._'
130	129	element_name = ''.join([

—
View it on GitLab.
You're receiving this email because of your account on gitlab.com. If you'd like to receive fewer emails, you can adjust your notification settings.