feat: massStoreRunAsynchronous()

Even though commit d915473 introduced a
socket-level TCP keepalive support into the server's implementation,
this was observed multiple times to not be enough to
**deterministically** fix the issues with the `CodeChecker store` client
hanging indefinitely when the server takes a long time processing the
to-be-stored data.
The underlying reasons are not entirely clear and the issue only pops
up sporadically, but we did observe a few similar scenarios (such as
multi-million report storage from analysing LLVM and then storing
between datacentres) where it almost reliably reproduces.
The symptoms (even with a configure `kepalive`) generally include the
server not becoming notified about the client's disconnect, while the
client process is hung on a low-level system call `read(4, ...)`, trying
to get the Thrift response of `massStoreRun()` from the HTTP socket.
Even if the server finishes the storage processing "in time" and sent
the Thrift reply, it never reaches the client, which means it never
exits from the waiting, which means it keeps either the terminal or,
worse, a CI script occupied, blocking execution.

This is the "more proper solution" foreshadowed in
commit 15af7d8.

Implemented the server-side logic to spawn a `MassStoreRun` task and
return its token, giving the `massStoreRunAsynchronous()` API call full
force.

Implemented the client-side logic to use the new `task_client` module
and the same logic as
`CodeChecker cmd serverside-tasks --await --token TOKEN...`
to poll the server for the task's completion and status.

docs/web/user_guide.md

@@ -480,6 +480,19 @@ optional arguments:
  is given, the longest match will be removed. You may
  also use Unix shell-like wildcards (e.g.
  '/*/jsmith/').
+ --detach Runs `store` in fire-and-forget mode: exit immediately
+ once the server accepted the analysis reports for
+ storing, without waiting for the server-side data
+ processing to conclude. Doing this is generally not
+ recommended, as the client will never be notified of
+ potential processing failures, and there is no easy way
+ to wait for the successfully stored results to become
+ available server-side for potential further processing
+ (e.g., `CodeChecker cmd diff`). However, using
+ '--detach' can significantly speed up large-scale
+ monitoring analyses where access to the results by a
+ tool is not a goal, such as in the case of non-gating
+ CI systems.
  --config CONFIG_FILE Allow the configuration from an explicit configuration
  file. The values configured in the config file will
  overwrite the values set in the command line.

web/api/codechecker_api_shared.thrift

@@ -34,13 +34,13 @@ enum ErrorCode {
  API_MISMATCH = 5,
 
  // REMOVED IN API v6.59 (CodeChecker v6.25.0)!
- // Previously sent by report_server.thrif/codeCheckerDBAccess::massStoreRun()
+ // Previously sent by report_server.thrift/codeCheckerDBAccess::massStoreRun()
  // when the client uploaded a source file which contained errors, such as
  // review status source-code-comment errors.
  /* SOURCE_FILE = 6, */ // Never reuse the value of the enum constant!
 
  // REMOVED IN API v6.59 (CodeChecker v6.25.0)!
- // Previously sent by report_server.thrif/codeCheckerDBAccess::massStoreRun()
+ // Previously sent by report_server.thrift/codeCheckerDBAccess::massStoreRun()
  // when the client uploaded a report with annotations that had invalid types.
  /* REPORT_FORMAT = 7, */ // Never reuse the value of the enum constant!
 }

web/client/codechecker_client/cmd/store.py

@@ -29,7 +29,8 @@
 from threading import Timer
 from typing import Dict, Iterable, List, Set, Tuple
 
-from codechecker_api.codeCheckerDBAccess_v6.ttypes import StoreLimitKind
+from codechecker_api.codeCheckerDBAccess_v6.ttypes import \
+ StoreLimitKind, SubmittedRunOptions
 
 from codechecker_report_converter import twodim
 from codechecker_report_converter.report import Report, report_file, \
@@ -48,8 +49,8 @@ def assemble_blame_info(_, __) -> int:
  """
  raise NotImplementedError()
 
-from codechecker_client import client as libclient
-from codechecker_client import product
+from codechecker_client import client as libclient, product
+from codechecker_client.task_client import await_task_termination
 from codechecker_common import arg, logger, cmd_config
 from codechecker_common.checker_labels import CheckerLabels
 from codechecker_common.compatibility.multiprocessing import Pool
@@ -254,6 +255,24 @@ def add_arguments_to_parser(parser):
  "match will be removed. You may also use Unix "
  "shell-like wildcards (e.g. '/*/jsmith/').")
 
+ parser.add_argument("--detach",
+ dest="detach",
+ default=argparse.SUPPRESS,
+ action="store_true",
+ required=False,
+ help="""
+Runs `store` in fire-and-forget mode: exit immediately once the server accepted
+the analysis reports for storing, without waiting for the server-side data
+processing to conclude.
+Doing this is generally not recommended, as the client will never be notified
+of potential processing failures, and there is no easy way to wait for the
+successfully stored results to become available server-side for potential
+further processing (e.g., `CodeChecker cmd diff`).
+However, using '--detach' can significantly speed up large-scale monitoring
+analyses where access to the results by a tool is not a goal, such as in the
+case of non-gating CI systems.
+""")
+
  cmd_config.add_option(parser)
 
  parser.add_argument('-f', '--force',
@@ -716,7 +735,7 @@ def get_analysis_statistics(inputs, limits):
  return statistics_files if has_failed_zip else []
 
 
-def storing_analysis_statistics(client, inputs, run_name):
+def store_analysis_statistics(client, inputs, run_name):
  """
  Collects and stores analysis statistics information on the server.
  """
@@ -931,68 +950,56 @@ def main(args):
 
  description = args.description if 'description' in args else None
 
- LOG.info("Storing results to the server...")
-
- try:
- with _timeout_watchdog(timedelta(hours=1),
- signal.SIGUSR1):
- client.massStoreRun(args.name,
- args.tag if 'tag' in args else None,
- str(context.version),
- b64zip,
- 'force' in args,
- trim_path_prefixes,
- description)
- except WatchdogError as we:
- LOG.warning("%s", str(we))
-
- # Showing parts of the exception stack is important here.
- # We **WANT** to see that the timeout happened during a wait on
- # Thrift reading from the TCP connection (something deep in the
- # Python library code at "sock.recv_into").
- import traceback
- _, _, tb = sys.exc_info()
- frames = traceback.extract_tb(tb)
- first, last = frames[0], frames[-2]
- formatted_frames = traceback.format_list([first, last])
- fmt_first, fmt_last = formatted_frames[0], formatted_frames[1]
- LOG.info("Timeout was triggered during:\n%s", fmt_first)
- LOG.info("Timeout interrupted this low-level operation:\n%s",
- fmt_last)
-
- LOG.error("Timeout!"
- "\n\tThe server's reply did not arrive after "
- "%d seconds (%s) elapsed since the server-side "
- "processing began."
- "\n\n\tThis does *NOT* mean that there was an issue "
- "with the run you were storing!"
- "\n\tThe server might still be processing the results..."
- "\n\tHowever, it is more likely that the "
- "server had already finished, but the client did not "
- "receive a response."
- "\n\tUsually, this is caused by the underlying TCP "
- "connection failing to signal a low-level disconnect."
- "\n\tClients potentially hanging indefinitely in these "
- "scenarios is an unfortunate and known issue."
- "\n\t\tSee http://github.com/Ericsson/codechecker/"
- "issues/3672 for details!"
- "\n\n\tThis error here is a temporary measure to ensure "
- "an infinite hang is replaced with a well-explained "
- "timeout."
- "\n\tA more proper solution will be implemented in a "
- "subsequent version of CodeChecker.",
- we.timeout.total_seconds(), str(we.timeout))
- sys.exit(1)
+ LOG.info("Storing results to the server ...")
+ task_token: str = client.massStoreRunAsynchronous(
+ b64zip,
+ SubmittedRunOptions(
+ runName=args.name,
+ tag=args.tag if "tag" in args else None,
+ version=str(context.version),
+ force="force" in args,
+ trimPathPrefixes=trim_path_prefixes,
+ description=description)
+ )
+ LOG.info("Reports submitted to the server for processing.")
 
- # Storing analysis statistics if the server allows them.
  if client.allowsStoringAnalysisStatistics():
- storing_analysis_statistics(client, args.input, args.name)
-
- LOG.info("Storage finished successfully.")
+ store_analysis_statistics(client, args.input, args.name)
+
+ if "detach" in args:
+ LOG.warning("Exiting the 'store' subcommand as '--detach' was "
+ "specified: not waiting for the result of the store "
+ "operation.\n"
+ "The server might not have finished processing "
+ "everything at this point, so do NOT rely on querying "
+ "the results just yet!\n"
+ "To await the completion of the processing later, "
+ "you can execute:\n\n"
+ "\tCodeChecker cmd serverside-tasks --token %s "
+ "--await",
+ task_token)
+ # Print the token to stdout as well, so scripts can use "--detach"
+ # meaningfully.
+ print(task_token)
+ return
+
+ task_client = libclient.setup_task_client(protocol, host, port)
+ task_status: str = await_task_termination(LOG, task_token,
+ task_api_client=task_client)
+
+ if task_status == "COMPLETED":
+ LOG.info("Storing the reports finished successfully.")
+ else:
+ LOG.error("Storing the reports failed! "
+ "The job terminated in status '%s'. "
+ "The comments associated with the failure are:\n\n%s",
+ task_status,
+ task_client.getTaskInfo(task_token).comments)
+ sys.exit(1)
  except Exception as ex:
  import traceback
  traceback.print_exc()
- LOG.info("Storage failed: %s", str(ex))
+ LOG.error("Storing the reports failed: %s", str(ex))
  sys.exit(1)
  finally:
  os.close(zip_file_handle)

web/client/codechecker_client/helpers/results.py

@@ -182,10 +182,12 @@ def massStoreRun(self, name, tag, version, zipdir, force,
  pass
 
  @thrift_client_call
- def massStoreRunAsynchronous(self, zipfile_blob: str,
- store_opts: ttypes.SubmittedRunOptions) \
- -> str:
- pass
+ def massStoreRunAsynchronous(
+ self,
+ zipfile_blob: str,
+ store_opts: ttypes.SubmittedRunOptions
+ ) -> str:
+ raise NotImplementedError("Should have called Thrift code!")
 
  @thrift_client_call
  def allowsStoringAnalysisStatistics(self):

web/client/codechecker_client/task_client.py

@@ -363,6 +363,7 @@ def _transform_product_ids_to_endpoints(
  product.id: product.endpoint
  for product
  in get_product_api().getProducts(None, None)}
+ ti["productEndpoint"] = product_id_to_endpoint[ti["productId"]]
  del ti["productId"]
 
 
@@ -445,7 +446,7 @@ def get_product_api() -> ThriftProductHelper:
  ti["taskKind"],
  ti["summary"],
  ti["status"],
- ti["productEndpoint"] or "",
+ ti.get("productEndpoint", ""),
  ti["actorUsername"] or "",
  ti["enqueuedAt"] or "",
  ti["startedAt"] or "",
@@ -529,7 +530,7 @@ def get_product_api() -> ThriftProductHelper:
  ti = task_info_for_print[0]
  product_line = \
  f" - Product: {ti['productEndpoint']}\n" \
- if ti["productEndpoint"] else ""
+ if "productEndpoint" in ti else ""
  user_line = f" - User: {ti['actorUsername']}\n" \
  if ti["actorUsername"] else ""
  cancel_line = " - Cancelled by administrators!\n" \

web/server/codechecker_server/api/mass_store_run.py

@@ -21,7 +21,8 @@
 import sqlalchemy
 import tempfile
 import time
-from typing import Any, Dict, List, Optional, Set, Tuple, Union, cast
+from typing import Any, Callable, Dict, List, Optional, Set, Tuple, Union, \
+ cast
 import zipfile
 import zlib
 
@@ -54,7 +55,7 @@
 from ..metadata import checker_is_unavailable, MetadataInfoParser
 from ..product import Product as ServerProduct
 from ..server_configuration import Configuration
-from ..task_executors.abstract_task import AbstractTask
+from ..task_executors.abstract_task import AbstractTask, TaskCancelHonoured
 from ..task_executors.task_manager import TaskManager
 from .thrift_enum_helper import report_extended_data_type_str
 
@@ -592,7 +593,12 @@ def _implementation(self, tm: TaskManager):
  f"'{self._product.endpoint}' is in "
  "a bad shape!")
 
- m = MassStoreRun(self.data_path / "store_zip",
+ def __cancel_if_needed():
+ if tm.should_cancel(self):
+ raise TaskCancelHonoured(self)
+
+ m = MassStoreRun(__cancel_if_needed,
+ self.data_path / "store_zip",
  self._package_context,
  tm.configuration_database_session_factory,
  self._product,
@@ -616,9 +622,8 @@ class MassStoreRun:
  # This is the place where complex implementation logic must go, but be
  # careful, there is no way to communicate with the user's client anymore!
 
- # TODO: Poll the task manager at regular points for a cancel signal!
-
  def __init__(self,
+ graceful_cancel: Callable[[], None],
  zip_dir: Path,
  package_context,
  config_db,
@@ -642,6 +647,7 @@ def __init__(self,
  self.__config_db = config_db
  self.__package_context = package_context
  self.__product = product
+ self.__graceful_cancel_if_requested = graceful_cancel
 
  self.__mips: Dict[str, MetadataInfoParser] = {}
  self.__analysis_info: Dict[str, AnalysisInfo] = {}
@@ -669,10 +675,10 @@ def __store_source_files(
  filename_to_hash: Dict[str, str]
  ) -> Dict[str, int]:
  """ Storing file contents from plist. """
-
  file_path_to_id = {}
 
  for file_name, file_hash in filename_to_hash.items():
+ self.__graceful_cancel_if_requested()
  source_file_path = path_for_fake_root(file_name, str(source_root))
  LOG.debug("Storing source file: %s", source_file_path)
  trimmed_file_path = trim_path_prefixes(
@@ -722,6 +728,7 @@ def __add_blame_info(
  with DBSession(self.__product.session_factory) as session:
  for subdir, _, files in os.walk(blame_root):
  for f in files:
+ self.__graceful_cancel_if_requested()
  blame_file = Path(subdir) / f
  file_path = f"/{str(blame_file.relative_to(blame_root))}"
  blame_info, remote_url, tracking_branch = \
@@ -1503,6 +1510,7 @@ def get_skip_handler(
  LOG.debug("Parsing input file '%s'", f)
 
  report_file_path = os.path.join(root_dir_path, f)
+ self.__graceful_cancel_if_requested()
  self.__process_report_file(
  report_file_path, session, run_id,
  file_path_to_id, run_history_time,
@@ -1604,6 +1612,7 @@ def store(self,
  original_zip_size: int,
  time_spent_on_task_preparation: float):
  """Store run results to the server."""
+ self.__graceful_cancel_if_requested()
  start_time = time.time()
 
  try:
@@ -1627,12 +1636,14 @@ def store(self,
 
  with StepLog(self._name, "Parse 'metadata.json's"):
  for root_dir_path, _, _ in os.walk(report_dir):
+ self.__graceful_cancel_if_requested()
  metadata_file_path = os.path.join(
  root_dir_path, 'metadata.json')
 
  self.__mips[root_dir_path] = \
  MetadataInfoParser(metadata_file_path)
 
+ self.__graceful_cancel_if_requested()
  with StepLog(self._name,
  "Store look-up ID for checkers in 'metadata.json'"):
  checkers_in_metadata = {
@@ -1656,10 +1667,16 @@ def store(self,
  session, report_dir, source_root, run_id,
  file_path_to_id, run_history_time)
 
+ self.__graceful_cancel_if_requested()
  session.commit()
  self.__load_report_ids_for_reports_with_fake_checkers(
  session)
 
+ # The task should not be cancelled after this point, as the
+ # "main" bulk of the modifications to the database had already
+ # been committed, and the user would be left with potentially
+ # a bunch of "fake checkers" visible in the database.
+
  if self.__reports_with_fake_checkers:
  with StepLog(
  self._name,
@@ -1720,6 +1737,8 @@ def store(self,
  LOG.error("Database error! Storing reports to the "
  "database failed: %s", ex)
  raise
+ except TaskCancelHonoured:
+ raise
  except Exception as ex:
  LOG.error("Failed to store results: %s", ex)
  import traceback