senders: Skips messages larger than 256K

GCP quotas page for the logging v2(https://cloud.google.com/logging/quotas) specifies
that a LogEntry should have at maximum 256KB in size, that is not a hard limit, but
an estimation.
In anycase we should check if any of our LogEntry objects are bigger than the quota
allows and not send those logs and instead let the user know what happened through a
custom message

Author: italomg

journalpump/senders/base.py

@@ -14,6 +14,8 @@
     import re  # type: ignore[no-redef]
 
 KAFKA_COMPRESSED_MESSAGE_OVERHEAD = 30
+
+# GCP logging also relies on this MAX message size
 MAX_KAFKA_MESSAGE_SIZE = 1024**2  # 1 MiB
 
 MAX_ERROR_MESSAGES = 8

journalpump/senders/google_cloud_logging.py

@@ -23,6 +23,11 @@ class GoogleCloudLoggingSender(LogSender):
         0: "EMERGENCY",
     }
 
+    # A bit on the safe side, not exactly 256KB but this
+    # is an approximation anyway
+    # according to https://cloud.google.com/logging/quotas
+    _LOG_ENTRY_QUOTA = 250 * 1024
+
     def __init__(self, *, config, googleapiclient_request_builder=None, **kwargs):
         super().__init__(config=config, max_send_interval=config.get("max_send_interval", 0.3), **kwargs)
         credentials = None
@@ -61,7 +66,14 @@ def send_messages(self, *, messages, cursor):
 
         for message in messages:
             msg_str = message.decode("utf8")
-            msg = json.loads(msg_str)
+            # This might not measure exactly 256K but should be
+            # a good enough approximation to handle this error.
+            # In anycase, we are handling this error in the
+            # exception handler below also.
+            if len(message) > GoogleCloudLoggingSender._LOG_ENTRY_QUOTA:
+                msg = { "MESSAGE": "Log entry cannot be logged because its size is greater than GCP logging quota of 256K" }
+            else:
+                msg = json.loads(msg_str)
             timestamp = msg.pop("timestamp", None)
             journald_priority = msg.pop("PRIORITY", None)
 

test/test_google_cloud_logging.py

@@ -144,3 +144,21 @@ def test_correct_timestamp(self):
             cursor=None,
         )
         assert sender._sent_count == 1  # pylint: disable=protected-access
+
+    def test_big_logentry_is_skip(self):
+        """Check that message was not marked as sent if GoogleApi returns error"""
+        request_builder = self._generate_request_builder(
+            [{"jsonPayload": { "MESSAGE": "Log entry cannot be logged because its size is greater than GCP logging quota of 256K" }}],
+        )
+
+        sender = GoogleCloudLoggingSender(
+            name="googlecloudlogging",
+            reader=mock.Mock(),
+            stats=mock.Mock(),
+            field_filter=None,
+            config=self.CONFIG,
+            googleapiclient_request_builder=request_builder,
+        )
+        message = f'{{"MESSAGE": "{"A" * 257_000}"}}'
+        sender.send_messages(messages=[message.encode()], cursor=None)
+        assert sender._sent_count == 1