def __init__(self) -> None:
columns = ColumnSet([
# columns to maintain the dataset
# Kafka topic offset
("offset", UInt(64)),
# GroupStatus in Sentry does not have a 'DELETED' state that reflects the deletion
# of the record. Having a dedicated clickhouse-only flag to identify this case seems
# more consistent than add an additional value into the status field below that does not
# exists on the Sentry side.
("record_deleted", UInt(8)),
# PG columns
("project_id", UInt(64)),
("id", UInt(64)),
("status", Nullable(UInt(8))),
("last_seen", Nullable(DateTime())),
("first_seen", Nullable(DateTime())),
("active_at", Nullable(DateTime())),
("first_release_id", Nullable(UInt(64))),
])
schema = ReplacingMergeTreeSchema(
columns=columns,
local_table_name="groupedmessage_local",
dist_table_name="groupedmessage_dist",
mandatory_conditions=[("record_deleted", "=", 0)],
prewhere_candidates=["project_id", "id"],
order_by="(project_id, id)",
partition_by=None,
version_column="offset",
sample_expr="id",
)
dataset_schemas = DatasetSchemas(
read_schema=schema,
write_schema=schema,
)
super().__init__(
dataset_schemas=dataset_schemas,
table_writer=GroupedMessageTableWriter(
write_schema=schema,
stream_loader=KafkaStreamLoader(
processor=GroupedMessageProcessor(self.POSTGRES_TABLE),
default_topic="cdc",
),
postgres_table=self.POSTGRES_TABLE,
),
default_control_topic="cdc_control",
postgres_table=self.POSTGRES_TABLE,
)
def __init__(self) -> None:
columns = ColumnSet([
# columns to maintain the dataset
# Kafka topic offset
("offset", UInt(64)),
("record_deleted", UInt(8)),
# PG columns
("project_id", UInt(64)),
("group_id", UInt(64)),
("date_added", Nullable(DateTime())),
("user_id", Nullable(UInt(64))),
("team_id", Nullable(UInt(64))),
])
schema = ReplacingMergeTreeSchema(
columns=columns,
local_table_name='groupassignee_local',
dist_table_name='groupassignee_dist',
order_by='(project_id, group_id)',
partition_by=None,
version_column='offset',
)
dataset_schemas = DatasetSchemas(
read_schema=schema,
write_schema=schema,
)
super().__init__(
dataset_schemas=dataset_schemas,
table_writer=GroupAssigneeTableWriter(
write_schema=schema,
stream_loader=KafkaStreamLoader(
processor=GroupAssigneeProcessor(self.POSTGRES_TABLE),
default_topic="cdc",
),
postgres_table=self.POSTGRES_TABLE,
),
default_control_topic="cdc_control",
postgres_table=self.POSTGRES_TABLE,
)
# during create statement
# (https://github.com/ClickHouse/ClickHouse/issues/12586), so the
# materialization is added with a migration.
skipped_cols_on_creation={"_tags_hash_map"},
)
storage = WritableTableStorage(
storage_key=StorageKey.TRANSACTIONS,
storage_set_key=StorageSetKey.TRANSACTIONS,
schema=schema,
query_processors=[
NestedFieldConditionOptimizer(
"contexts",
"_contexts_flattened",
{"start_ts", "finish_ts"},
BEGINNING_OF_TIME,
),
MappingOptimizer("tags", "_tags_hash_map", "tags_hash_map_enabled"),
TransactionColumnProcessor(),
ArrayJoinKeyValueOptimizer("tags"),
ArrayJoinKeyValueOptimizer("measurements"),
PrewhereProcessor(),
],
stream_loader=KafkaStreamLoader(
processor=TransactionsMessageProcessor(), default_topic="events",
),
query_splitters=[TimeSplitQueryStrategy(timestamp_col="finish_ts")],
writer_options={"insert_allow_materialized_columns": 1},
)
"deleted",
"retention_days",
]
storage = WritableTableStorage(
storage_key=StorageKey.ERRORS,
storage_set_key=StorageSetKey.EVENTS,
schema=schema,
query_processors=[
PostReplacementConsistencyEnforcer(
project_column="project_id", replacer_state_name=ReplacerState.ERRORS,
),
MappingColumnPromoter(mapping_specs={"tags": promoted_tag_columns}),
ArrayJoinKeyValueOptimizer("tags"),
PrewhereProcessor(),
],
stream_loader=KafkaStreamLoader(
processor=ErrorsProcessor(promoted_tag_columns),
default_topic="events",
replacement_topic="errors-replacements",
),
replacer_processor=ErrorsReplacer(
write_schema=schema,
read_schema=schema,
required_columns=required_columns,
tag_column_map={"tags": promoted_tag_columns, "contexts": {}},
promoted_tags={"tags": list(promoted_tag_columns.keys()), "contexts": []},
state_name=ReplacerState.ERRORS,
),
)
None,
ConditionFunctions.EQ,
Column(None, None, "record_deleted"),
Literal(None, 0),
),
)
],
prewhere_candidates=["project_id", "id"],
order_by="(project_id, id)",
partition_by=None,
version_column="offset",
sample_expr="id",
)
POSTGRES_TABLE = "sentry_groupedmessage"
storage = CdcStorage(
storage_key=StorageKey.GROUPEDMESSAGES,
storage_set_key=StorageSetKey.EVENTS,
schema=schema,
query_processors=[],
stream_loader=KafkaStreamLoader(
processor=GroupedMessageProcessor(POSTGRES_TABLE),
default_topic="cdc",
pre_filter=CdcTableNameMessageFilter(POSTGRES_TABLE),
),
default_control_topic="cdc_control",
postgres_table=POSTGRES_TABLE,
row_processor=lambda row: GroupedMessageRow.from_bulk(row).to_clickhouse(),
)
schema = WritableTableSchema(
columns=all_columns,
local_table_name="sentry_local",
dist_table_name="sentry_dist",
storage_set_key=StorageSetKey.EVENTS,
mandatory_conditions=mandatory_conditions,
prewhere_candidates=prewhere_candidates,
)
storage = WritableTableStorage(
storage_key=StorageKey.EVENTS,
storage_set_key=StorageSetKey.EVENTS,
schema=schema,
query_processors=query_processors,
stream_loader=KafkaStreamLoader(
processor=EventsProcessor(promoted_tag_columns),
default_topic="events",
replacement_topic="event-replacements",
commit_log_topic="snuba-commit-log",
),
query_splitters=query_splitters,
replacer_processor=ErrorsReplacer(
schema=schema,
required_columns=[col.escaped for col in required_columns],
tag_column_map=get_tag_column_map(),
promoted_tags=get_promoted_tags(),
state_name=ReplacerState.EVENTS,
),
)
def __init__(self) -> None:
columns = ColumnSet(
[
("project_id", UInt(64)),
("event_id", UUID()),
("trace_id", UUID()),
("span_id", UInt(64)),
("transaction_name", LowCardinality(String())),
(
"transaction_hash",
Materialized(UInt(64), "cityHash64(transaction_name)",),
),
("transaction_op", LowCardinality(String())),
("transaction_status", WithDefault(UInt(8), UNKNOWN_SPAN_STATUS)),
("start_ts", DateTime()),
("start_ms", UInt(16)),
("_start_date", Materialized(Date(), "toDate(start_ts)"),),
("finish_ts", DateTime()),
("finish_ms", UInt(16)),
("_finish_date", Materialized(Date(), "toDate(finish_ts)"),),
("duration", UInt(32)),
("platform", LowCardinality(String())),
("environment", LowCardinality(Nullable(String()))),
("release", LowCardinality(Nullable(String()))),
("dist", LowCardinality(Nullable(String()))),
("ip_address_v4", Nullable(IPv4())),
("ip_address_v6", Nullable(IPv6())),
("user", WithDefault(String(), "''",)),
("user_hash", Materialized(UInt(64), "cityHash64(user)"),),
("user_id", Nullable(String())),
("user_name", Nullable(String())),
("user_email", Nullable(String())),
("sdk_name", WithDefault(LowCardinality(String()), "''")),
("sdk_version", WithDefault(LowCardinality(String()), "''")),
("tags", Nested([("key", String()), ("value", String())])),
("_tags_flattened", String()),
("contexts", Nested([("key", String()), ("value", String())])),
("_contexts_flattened", String()),
("partition", UInt(16)),
("offset", UInt(64)),
("retention_days", UInt(16)),
("deleted", UInt(8)),
]
)
schema = ReplacingMergeTreeSchema(
columns=columns,
local_table_name="transactions_local",
dist_table_name="transactions_dist",
mandatory_conditions=[],
prewhere_candidates=["event_id", "project_id"],
order_by="(project_id, _finish_date, transaction_name, cityHash64(span_id))",
partition_by="(retention_days, toMonday(_finish_date))",
version_column="deleted",
sample_expr=None,
migration_function=transactions_migrations,
)
dataset_schemas = DatasetSchemas(read_schema=schema, write_schema=schema,)
self.__tags_processor = TagColumnProcessor(
columns=columns,
promoted_columns=self._get_promoted_columns(),
column_tag_map=self._get_column_tag_map(),
)
super().__init__(
dataset_schemas=dataset_schemas,
table_writer=TransactionsTableWriter(
write_schema=schema,
stream_loader=KafkaStreamLoader(
processor=TransactionsMessageProcessor(), default_topic="events",
),
),
time_group_columns={
"bucketed_start": "start_ts",
"bucketed_end": "finish_ts",
},
time_parse_columns=("start_ts", "finish_ts"),
)
storage_set_key=StorageSetKey.SESSIONS,
prewhere_candidates=["project_id", "org_id"],
)
materialized_view_schema = TableSchema(
local_table_name=READ_LOCAL_MV_NAME,
dist_table_name=READ_DIST_MV_NAME,
storage_set_key=StorageSetKey.SESSIONS,
prewhere_candidates=["project_id", "org_id"],
columns=read_columns,
)
# The raw table we write onto, and that potentially we could
# query.
raw_storage = WritableTableStorage(
storage_key=StorageKey.SESSIONS_RAW,
storage_set_key=StorageSetKey.SESSIONS,
schema=raw_schema,
query_processors=[],
stream_loader=KafkaStreamLoader(
processor=SessionsProcessor(),
default_topic="ingest-sessions",
),
)
# The materialized view we query aggregate data from.
materialized_storage = ReadableTableStorage(
storage_key=StorageKey.SESSIONS_HOURLY,
storage_set_key=StorageSetKey.SESSIONS,
schema=read_schema,
query_processors=[PrewhereProcessor()],
)
def __init__(self):
columns = ColumnSet([
('project_id', UInt(64)),
('event_id', UUID()),
('trace_id', UUID()),
('span_id', UInt(64)),
('transaction_name', String()),
('transaction_hash',
Materialized(
UInt(64),
'cityHash64(transaction_name)',
)),
('transaction_op', LowCardinality(String())),
('start_ts', DateTime()),
('start_ms', UInt(16)),
('finish_ts', DateTime()),
('finish_ms', UInt(16)),
('duration', UInt(32)),
('platform', LowCardinality(String())),
('environment', Nullable(String())),
('release', Nullable(String())),
('dist', Nullable(String())),
('ip_address_v4', Nullable(IPv4())),
('ip_address_v6', Nullable(IPv6())),
('user', WithDefault(
String(),
"''",
)),
('user_id', Nullable(String())),
('user_name', Nullable(String())),
('user_email', Nullable(String())),
('tags', Nested([
('key', String()),
('value', String()),
])),
('contexts', Nested([
('key', String()),
('value', String()),
])),
('partition', UInt(16)),
('offset', UInt(64)),
('retention_days', UInt(16)),
('deleted', UInt(8)),
])
schema = ReplacingMergeTreeSchema(
columns=columns,
local_table_name='transactions_local',
dist_table_name='transactions_dist',
mandatory_conditions=[],
order_by=
'(project_id, toStartOfDay(start_ts), transaction_hash, start_ts, start_ms, trace_id, span_id)',
partition_by='(retention_days, toMonday(start_ts))',
version_column='deleted',
sample_expr=None,
migration_function=transactions_migrations,
)
dataset_schemas = DatasetSchemas(
read_schema=schema,
write_schema=schema,
)
self.__tags_processor = TagColumnProcessor(
columns=columns,
promoted_columns=self._get_promoted_columns(),
column_tag_map=self._get_column_tag_map(),
)
super().__init__(dataset_schemas=dataset_schemas,
table_writer=TransactionsTableWriter(
write_schema=schema,
stream_loader=KafkaStreamLoader(
processor=TransactionsMessageProcessor(),
default_topic="events",
),
),
time_group_columns={
'bucketed_start': 'start_ts',
'bucketed_end': 'finish_ts',
},
time_parse_columns=('start_ts', 'finish_ts'))
("max_threads", UInt(8)),
("num_days", UInt(32)),
("clickhouse_table", LowCardinality(String())),
("query_id", String()),
("is_duplicate", UInt(8)),
("consistent", UInt(8)),
]),
),
])
schema = MergeTreeSchema(
columns=columns,
local_table_name="querylog_local",
dist_table_name="querylog_dist",
order_by="(toStartOfDay(timestamp), request_id)",
partition_by="(toMonday(timestamp))",
sample_expr="request_id",
)
storage = WritableTableStorage(
schemas=StorageSchemas(read_schema=schema, write_schema=schema),
table_writer=TableWriter(
write_schema=schema,
stream_loader=KafkaStreamLoader(
processor=QuerylogProcessor(),
default_topic=settings.QUERIES_TOPIC,
),
),
query_processors=[],
)
def __init__(self) -> None:
write_columns = ColumnSet([
("org_id", UInt(64)),
("project_id", UInt(64)),
("key_id", Nullable(UInt(64))),
("timestamp", DateTime()),
("outcome", UInt(8)),
("reason", LowCardinality(Nullable(String()))),
("event_id", Nullable(UUID())),
])
write_schema = MergeTreeSchema(
columns=write_columns,
# TODO: change to outcomes.raw_local when we add multi DB support
local_table_name=WRITE_LOCAL_TABLE_NAME,
dist_table_name=WRITE_DIST_TABLE_NAME,
order_by="(org_id, project_id, timestamp)",
partition_by="(toMonday(timestamp))",
settings={"index_granularity": 16384},
)
read_columns = ColumnSet([
("org_id", UInt(64)),
("project_id", UInt(64)),
("key_id", UInt(64)),
("timestamp", DateTime()),
("outcome", UInt(8)),
("reason", LowCardinality(String())),
("times_seen", UInt(64)),
])
read_schema = SummingMergeTreeSchema(
columns=read_columns,
local_table_name=READ_LOCAL_TABLE_NAME,
dist_table_name=READ_DIST_TABLE_NAME,
order_by="(org_id, project_id, key_id, outcome, reason, timestamp)",
partition_by="(toMonday(timestamp))",
settings={"index_granularity": 256},
)
materialized_view_columns = ColumnSet([
("org_id", UInt(64)),
("project_id", UInt(64)),
("key_id", UInt(64)),
("timestamp", DateTime()),
("outcome", UInt(8)),
("reason", String()),
("times_seen", UInt(64)),
])
# TODO: Find a better way to specify a query for a materialized view
# The problem right now is that we have a way to define our columns in a ColumnSet abstraction but the query
# doesn't use it.
query = """
SELECT
org_id,
project_id,
ifNull(key_id, 0) AS key_id,
toStartOfHour(timestamp) AS timestamp,
outcome,
ifNull(reason, 'none') AS reason,
count() AS times_seen
FROM %(source_table_name)s
GROUP BY org_id, project_id, key_id, timestamp, outcome, reason
"""
materialized_view = MaterializedViewSchema(
local_materialized_view_name="outcomes_mv_hourly_local",
dist_materialized_view_name="outcomes_mv_hourly_dist",
prewhere_candidates=["project_id", "org_id"],
columns=materialized_view_columns,
query=query,
local_source_table_name=WRITE_LOCAL_TABLE_NAME,
local_destination_table_name=READ_LOCAL_TABLE_NAME,
dist_source_table_name=WRITE_DIST_TABLE_NAME,
dist_destination_table_name=READ_DIST_TABLE_NAME,
)
dataset_schemas = DatasetSchemas(
read_schema=read_schema,
write_schema=write_schema,
intermediary_schemas=[materialized_view],
)
table_writer = TableWriter(
write_schema=write_schema,
stream_loader=KafkaStreamLoader(
processor=OutcomesProcessor(),
default_topic="outcomes",
),
)
super().__init__(
dataset_schemas=dataset_schemas,
table_writer=table_writer,
time_group_columns={"time": "timestamp"},
time_parse_columns=("timestamp", ),
)
def __init__(self):
metadata_columns = ColumnSet([
# optional stream related data
('offset', Nullable(UInt(64))),
('partition', Nullable(UInt(16))),
])
promoted_tag_columns = ColumnSet([
# These are the classic tags, they are saved in Snuba exactly as they
# appear in the event body.
('level', Nullable(String())),
('logger', Nullable(String())),
('server_name', Nullable(String())), # future name: device_id?
('transaction', Nullable(String())),
('environment', Nullable(String())),
('sentry:release', Nullable(String())),
('sentry:dist', Nullable(String())),
('sentry:user', Nullable(String())),
('site', Nullable(String())),
('url', Nullable(String())),
])
promoted_context_tag_columns = ColumnSet([
# These are promoted tags that come in in `tags`, but are more closely
# related to contexts. To avoid naming confusion with Clickhouse nested
# columns, they are stored in the database with s/./_/
# promoted tags
('app_device', Nullable(String())),
('device', Nullable(String())),
('device_family', Nullable(String())),
('runtime', Nullable(String())),
('runtime_name', Nullable(String())),
('browser', Nullable(String())),
('browser_name', Nullable(String())),
('os', Nullable(String())),
('os_name', Nullable(String())),
('os_rooted', Nullable(UInt(8))),
])
promoted_context_columns = ColumnSet([
('os_build', Nullable(String())),
('os_kernel_version', Nullable(String())),
('device_name', Nullable(String())),
('device_brand', Nullable(String())),
('device_locale', Nullable(String())),
('device_uuid', Nullable(String())),
('device_model_id', Nullable(String())),
('device_arch', Nullable(String())),
('device_battery_level', Nullable(Float(32))),
('device_orientation', Nullable(String())),
('device_simulator', Nullable(UInt(8))),
('device_online', Nullable(UInt(8))),
('device_charging', Nullable(UInt(8))),
])
required_columns = ColumnSet([
('event_id', FixedString(32)),
('project_id', UInt(64)),
('group_id', UInt(64)),
('timestamp', DateTime()),
('deleted', UInt(8)),
('retention_days', UInt(16)),
])
all_columns = required_columns + [
# required for non-deleted
('platform', Nullable(String())),
('message', Nullable(String())),
('primary_hash', Nullable(FixedString(32))),
('received', Nullable(DateTime())),
('search_message', Nullable(String())),
('title', Nullable(String())),
('location', Nullable(String())),
# optional user
('user_id', Nullable(String())),
('username', Nullable(String())),
('email', Nullable(String())),
('ip_address', Nullable(String())),
# optional geo
('geo_country_code', Nullable(String())),
('geo_region', Nullable(String())),
('geo_city', Nullable(String())),
('sdk_name', Nullable(String())),
('sdk_version', Nullable(String())),
('type', Nullable(String())),
('version', Nullable(String())),
] + metadata_columns \
+ promoted_context_columns \
+ promoted_tag_columns \
+ promoted_context_tag_columns \
+ [
# other tags
('tags', Nested([
('key', String()),
('value', String()),
])),
# other context
('contexts', Nested([
('key', String()),
('value', String()),
])),
# http interface
('http_method', Nullable(String())),
('http_referer', Nullable(String())),
# exception interface
('exception_stacks', Nested([
('type', Nullable(String())),
('value', Nullable(String())),
('mechanism_type', Nullable(String())),
('mechanism_handled', Nullable(UInt(8))),
])),
('exception_frames', Nested([
('abs_path', Nullable(String())),
('filename', Nullable(String())),
('package', Nullable(String())),
('module', Nullable(String())),
('function', Nullable(String())),
('in_app', Nullable(UInt(8))),
('colno', Nullable(UInt(32))),
('lineno', Nullable(UInt(32))),
('stack_level', UInt(16)),
])),
# These are columns we added later in the life of the (current) production
# database. They don't necessarily belong here in a logical/readability sense
# but they are here to match the order of columns in production becase
# `insert_distributed_sync` is very sensitive to column existence and ordering.
('culprit', Nullable(String())),
('sdk_integrations', Array(String())),
('modules', Nested([
('name', String()),
('version', String()),
])),
]
sample_expr = 'cityHash64(toString(event_id))'
schema = ReplacingMergeTreeSchema(
columns=all_columns,
local_table_name='sentry_local',
dist_table_name='sentry_dist',
mandatory_conditions=[('deleted', '=', 0)],
order_by='(project_id, toStartOfDay(timestamp), %s)' % sample_expr,
partition_by='(toMonday(timestamp), if(equals(retention_days, 30), 30, 90))',
version_column='deleted',
sample_expr=sample_expr,
migration_function=events_migrations)
dataset_schemas = DatasetSchemas(
read_schema=schema,
write_schema=schema,
)
table_writer = TableWriter(
write_schema=schema,
stream_loader=KafkaStreamLoader(
processor=EventsProcessor(promoted_tag_columns),
default_topic="events",
replacement_topic="event-replacements",
commit_log_topic="snuba-commit-log",
)
)
super(EventsDataset, self).__init__(
dataset_schemas=dataset_schemas,
table_writer=table_writer,
time_group_columns={
'time': 'timestamp',
'rtime': 'received'
},
time_parse_columns=('timestamp', 'received')
)
self.__metadata_columns = metadata_columns
self.__promoted_tag_columns = promoted_tag_columns
self.__promoted_context_tag_columns = promoted_context_tag_columns
self.__promoted_context_columns = promoted_context_columns
self.__required_columns = required_columns
self.__tags_processor = TagColumnProcessor(
columns=all_columns,
promoted_columns=self._get_promoted_columns(),
column_tag_map=self._get_column_tag_map(),
)
materialized_view_schema = MaterializedViewSchema(
local_materialized_view_name="outcomes_mv_hourly_local",
dist_materialized_view_name="outcomes_mv_hourly_dist",
storage_set_key=StorageSetKey.OUTCOMES,
prewhere_candidates=["project_id", "org_id"],
columns=materialized_view_columns,
query=query,
local_source_table_name=WRITE_LOCAL_TABLE_NAME,
local_destination_table_name=READ_LOCAL_TABLE_NAME,
dist_source_table_name=WRITE_DIST_TABLE_NAME,
dist_destination_table_name=READ_DIST_TABLE_NAME,
)
raw_storage = WritableTableStorage(
storage_key=StorageKey.OUTCOMES_RAW,
storage_set_key=StorageSetKey.OUTCOMES,
schema=raw_schema,
query_processors=[],
stream_loader=KafkaStreamLoader(
processor=OutcomesProcessor(),
default_topic="outcomes",
),
)
materialized_storage = ReadableTableStorage(
storage_key=StorageKey.OUTCOMES_HOURLY,
storage_set_key=StorageSetKey.OUTCOMES,
schema=read_schema,
query_processors=[PrewhereProcessor()],
)
def __init__(self) -> None:
metadata_columns = ColumnSet([
# optional stream related data
("offset", Nullable(UInt(64))),
("partition", Nullable(UInt(16))),
])
promoted_tag_columns = ColumnSet([
# These are the classic tags, they are saved in Snuba exactly as they
# appear in the event body.
("level", Nullable(String())),
("logger", Nullable(String())),
("server_name", Nullable(String())), # future name: device_id?
("transaction", Nullable(String())),
("environment", Nullable(String())),
("sentry:release", Nullable(String())),
("sentry:dist", Nullable(String())),
("sentry:user", Nullable(String())),
("site", Nullable(String())),
("url", Nullable(String())),
])
promoted_context_tag_columns = ColumnSet([
# These are promoted tags that come in in `tags`, but are more closely
# related to contexts. To avoid naming confusion with Clickhouse nested
# columns, they are stored in the database with s/./_/
# promoted tags
("app_device", Nullable(String())),
("device", Nullable(String())),
("device_family", Nullable(String())),
("runtime", Nullable(String())),
("runtime_name", Nullable(String())),
("browser", Nullable(String())),
("browser_name", Nullable(String())),
("os", Nullable(String())),
("os_name", Nullable(String())),
("os_rooted", Nullable(UInt(8))),
])
promoted_context_columns = ColumnSet([
("os_build", Nullable(String())),
("os_kernel_version", Nullable(String())),
("device_name", Nullable(String())),
("device_brand", Nullable(String())),
("device_locale", Nullable(String())),
("device_uuid", Nullable(String())),
("device_model_id", Nullable(String())),
("device_arch", Nullable(String())),
("device_battery_level", Nullable(Float(32))),
("device_orientation", Nullable(String())),
("device_simulator", Nullable(UInt(8))),
("device_online", Nullable(UInt(8))),
("device_charging", Nullable(UInt(8))),
])
required_columns = ColumnSet([
("event_id", FixedString(32)),
("project_id", UInt(64)),
("group_id", UInt(64)),
("timestamp", DateTime()),
("deleted", UInt(8)),
("retention_days", UInt(16)),
])
all_columns = (
required_columns + [
# required for non-deleted
("platform", Nullable(String())),
("message", Nullable(String())),
("primary_hash", Nullable(FixedString(32))),
("received", Nullable(DateTime())),
("search_message", Nullable(String())),
("title", Nullable(String())),
("location", Nullable(String())),
# optional user
("user_id", Nullable(String())),
("username", Nullable(String())),
("email", Nullable(String())),
("ip_address", Nullable(String())),
# optional geo
("geo_country_code", Nullable(String())),
("geo_region", Nullable(String())),
("geo_city", Nullable(String())),
("sdk_name", Nullable(String())),
("sdk_version", Nullable(String())),
("type", Nullable(String())),
("version", Nullable(String())),
] + metadata_columns + promoted_context_columns +
promoted_tag_columns + promoted_context_tag_columns + [
# other tags
("tags", Nested([("key", String()), ("value", String())])),
("_tags_flattened", String()),
# other context
("contexts", Nested([("key", String()), ("value", String())])),
# http interface
("http_method", Nullable(String())),
("http_referer", Nullable(String())),
# exception interface
(
"exception_stacks",
Nested([
("type", Nullable(String())),
("value", Nullable(String())),
("mechanism_type", Nullable(String())),
("mechanism_handled", Nullable(UInt(8))),
]),
),
(
"exception_frames",
Nested([
("abs_path", Nullable(String())),
("filename", Nullable(String())),
("package", Nullable(String())),
("module", Nullable(String())),
("function", Nullable(String())),
("in_app", Nullable(UInt(8))),
("colno", Nullable(UInt(32))),
("lineno", Nullable(UInt(32))),
("stack_level", UInt(16)),
]),
),
# These are columns we added later in the life of the (current) production
# database. They don't necessarily belong here in a logical/readability sense
# but they are here to match the order of columns in production becase
# `insert_distributed_sync` is very sensitive to column existence and ordering.
("culprit", Nullable(String())),
("sdk_integrations", Array(String())),
("modules", Nested([("name", String()),
("version", String())])),
])
sample_expr = "cityHash64(toString(event_id))"
schema = ReplacingMergeTreeSchema(
columns=all_columns,
local_table_name="sentry_local",
dist_table_name="sentry_dist",
mandatory_conditions=[("deleted", "=", 0)],
prewhere_candidates=[
"event_id",
"group_id",
"tags[sentry:release]",
"message",
"environment",
"project_id",
],
order_by="(project_id, toStartOfDay(timestamp), %s)" % sample_expr,
partition_by=
"(toMonday(timestamp), if(equals(retention_days, 30), 30, 90))",
version_column="deleted",
sample_expr=sample_expr,
migration_function=events_migrations,
)
dataset_schemas = DatasetSchemas(
read_schema=schema,
write_schema=schema,
)
table_writer = TableWriter(
write_schema=schema,
stream_loader=KafkaStreamLoader(
processor=EventsProcessor(promoted_tag_columns),
default_topic="events",
replacement_topic="event-replacements",
commit_log_topic="snuba-commit-log",
),
)
super(EventsDataset, self).__init__(
dataset_schemas=dataset_schemas,
table_writer=table_writer,
time_group_columns={
"time": "timestamp",
"rtime": "received"
},
time_parse_columns=("timestamp", "received"),
)
self.__metadata_columns = metadata_columns
self.__promoted_tag_columns = promoted_tag_columns
self.__promoted_context_tag_columns = promoted_context_tag_columns
self.__promoted_context_columns = promoted_context_columns
self.__required_columns = required_columns
self.__tags_processor = TagColumnProcessor(
columns=all_columns,
promoted_columns=self._get_promoted_columns(),
column_tag_map=self._get_column_tag_map(),
)
def __init__(self) -> None:
all_columns = ColumnSet([
("org_id", UInt(64)),
("project_id", UInt(64)),
("timestamp", DateTime()),
("event_id", WithCodecs(UUID(), ["NONE"])),
(
"event_hash",
WithCodecs(
Materialized(
UInt(64),
"cityHash64(toString(event_id))",
),
["NONE"],
),
),
("platform", LowCardinality(String())),
("environment", LowCardinality(Nullable(String()))),
("release", LowCardinality(Nullable(String()))),
("dist", LowCardinality(Nullable(String()))),
("ip_address_v4", Nullable(IPv4())),
("ip_address_v6", Nullable(IPv6())),
("user", WithDefault(String(), "''")),
(
"user_hash",
Materialized(UInt(64), "cityHash64(user)"),
),
("user_id", Nullable(String())),
("user_name", Nullable(String())),
("user_email", Nullable(String())),
("sdk_name", LowCardinality(Nullable(String()))),
("sdk_version", LowCardinality(Nullable(String()))),
("tags", Nested([("key", String()), ("value", String())])),
("_tags_flattened", String()),
("contexts", Nested([("key", String()), ("value", String())])),
("_contexts_flattened", String()),
("transaction_name", WithDefault(LowCardinality(String()), "''")),
(
"transaction_hash",
Materialized(UInt(64), "cityHash64(transaction_name)"),
),
("span_id", Nullable(UInt(64))),
("trace_id", Nullable(UUID())),
("partition", UInt(16)),
("offset", WithCodecs(UInt(64), ["DoubleDelta", "LZ4"])),
("retention_days", UInt(16)),
("deleted", UInt(8)),
("group_id", UInt(64)),
("primary_hash", FixedString(32)),
("primary_hash_hex", Materialized(UInt(64), "hex(primary_hash)")),
("event_string", WithCodecs(String(), ["NONE"])),
("received", DateTime()),
("message", String()),
("title", String()),
("culprit", String()),
("level", LowCardinality(String())),
("location", Nullable(String())),
("version", LowCardinality(Nullable(String()))),
("type", LowCardinality(String())),
(
"exception_stacks",
Nested([
("type", Nullable(String())),
("value", Nullable(String())),
("mechanism_type", Nullable(String())),
("mechanism_handled", Nullable(UInt(8))),
]),
),
(
"exception_frames",
Nested([
("abs_path", Nullable(String())),
("colno", Nullable(UInt(32))),
("filename", Nullable(String())),
("function", Nullable(String())),
("lineno", Nullable(UInt(32))),
("in_app", Nullable(UInt(8))),
("package", Nullable(String())),
("module", Nullable(String())),
("stack_level", Nullable(UInt(16))),
]),
),
("sdk_integrations", Array(String())),
("modules", Nested([("name", String()), ("version", String())])),
])
self.__promoted_tag_columns = {
"environment": "environment",
"sentry:release": "release",
"sentry:dist": "dist",
"sentry:user": "******",
"transaction": "transaction_name",
"level": "level",
}
schema = ReplacingMergeTreeSchema(
columns=all_columns,
local_table_name="errors_local",
dist_table_name="errors_dist",
mandatory_conditions=[("deleted", "=", 0)],
prewhere_candidates=[
"event_id",
"group_id",
"tags[sentry:release]",
"message",
"environment",
"project_id",
],
order_by=
"(org_id, project_id, toStartOfDay(timestamp), primary_hash_hex, event_hash)",
partition_by=
"(toMonday(timestamp), if(retention_days = 30, 30, 90))",
version_column="deleted",
sample_expr="event_hash",
ttl_expr="timestamp + toIntervalDay(retention_days)",
settings={"index_granularity": "8192"},
)
dataset_schemas = DatasetSchemas(
read_schema=schema,
write_schema=schema,
)
table_writer = TableWriter(
write_schema=schema,
stream_loader=KafkaStreamLoader(
processor=ErrorsProcessor(self.__promoted_tag_columns),
default_topic="events",
),
)
super().__init__(
dataset_schemas=dataset_schemas,
table_writer=table_writer,
time_group_columns={
"time": "timestamp",
"rtime": "received"
},
time_parse_columns=("timestamp", "received"),
)
self.__tags_processor = TagColumnProcessor(
columns=all_columns,
promoted_columns=self._get_promoted_columns(),
column_tag_map=self._get_column_tag_map(),
)