I was just reviewing the documentation to understand how Google Dataflow handles watermarks, and it just mentions the very vague:
The data source determines the watermark
It seems you can add more flexibility through withAllowedLateness but what will happen if we do not configure this?
Thoughts so far
I found something indicating that if your source is Google PubSub it already has a watermark which will get taken, but what if the source is something else? For example a Kafka topic (which I believe does not inherently have a watermark, so I don't see how something like this would apply).
Is it always 10 seconds, or just 0? Is it looking at the last few minutes to determine the max lag and if so how many (surely not since forever as that would get distorted by the initial start of processing which might see giant lag)? I could not find anything on the topic.
I also searched outside the context of Google DataFlow for Apache Beam documentation but did not find anything explaining this either.
When using Apache Kafka as a data source, each Kafka partition may have a simple event time pattern (ascending timestamps or bounded out-of-orderness). However, when consuming streams from Kafka, multiple partitions often get consumed in parallel, interleaving the events from the partitions and destroying the per-partition patterns (this is inherent in how Kafka’s consumer clients work).
In that case, you can use Flink’s Kafka-partition-aware watermark generation. Using that feature, watermarks are generated inside the Kafka consumer, per Kafka partition, and the per-partition watermarks are merged in the same way as watermarks are merged on stream shuffles.
For example, if event timestamps are strictly ascending per Kafka partition, generating per-partition watermarks with the ascending timestamps watermark generator will result in perfect overall watermarks. Note, that TimestampAssigner is not provided in the example, the timestamps of the Kafka records themselves will be used instead.
In any data processing system, there is a certain amount of lag between the time a data event occurs (the “event time”, determined by the timestamp on the data element itself) and the time the actual data element gets processed at any stage in your pipeline (the “processing time”, determined by the clock on the system processing the element). In addition, there are no guarantees that data events will appear in your pipeline in the same order that they were generated.
For example, let’s say we have a PCollection that’s using fixed-time windowing, with windows that are five minutes long. For each window, Beam must collect all the data with an event time timestamp in the given window range (between 0:00 and 4:59 in the first window, for instance). Data with timestamps outside that range (data from 5:00 or later) belong to a different window.
However, data isn’t always guaranteed to arrive in a pipeline in time order, or to always arrive at predictable intervals. Beam tracks a watermark, which is the system’s notion of when all data in a certain window can be expected to have arrived in the pipeline. Once the watermark progresses past the end of a window, any further element that arrives with a timestamp in that window is considered late data.
From our example, suppose we have a simple watermark that assumes approximately 30s of lag time between the data timestamps (the event time) and the time the data appears in the pipeline (the processing time), then Beam would close the first window at 5:30. If a data record arrives at 5:34, but with a timestamp that would put it in the 0:00-4:59 window (say, 3:38), then that record is late data.
Related
I'm using Cassandra and Kafka for event-sourcing, and it works quite well. But I've just recently discovered a potentially major flaw in the design/set-up. A brief intro to how it is done:
The aggregate command handler is basically a kafka consumer, which consumes messages of interest on a topic:
1.1 When it receives a command, it loads all events for the aggregate, and replays the aggregate event handler for each event to get the aggregate up to current state.
1.2 Based on the command and businiss logic it then applies one or more events to the event store. This involves inserting the new event(s) to the event store table in cassandra. The events are stamped with a version number for the aggregate - starting at version 0 for a new aggregate, making projections possible. In addition it sends the event to another topic (for projection purposes).
1.3 A kafka consumer will listen on the topic upon these events are published. This consumer will act as a projector. When it receives an event of interest, it loads the current read model for the aggregate. It checks that the version of the event it has received is the expected version, and then updates the read model.
This seems to work very well. The problem is when I want to have what EventStore calls category projections. Let's take Order aggregate as an example. I can easily project one or more read models pr Order. But if I want to for example have a projection which contains a customers 30 last orders, then I would need a category projection.
I'm just scratching my head how to accomplish this. I'm curious to know if any other are using Cassandra and Kafka for event sourcing. I've read a couple of places that some people discourage it. Maybe this is the reason.
I know EventStore has support for this built in. Maybe using Kafka as event store would be a better solution.
With this kind of architecture, you have to choose between:
Global event stream per type - simple
Partitioned event stream per type - scalable
Unless your system is fairly high throughput (say at least 10s or 100s of events per second for sustained periods to the stream type in question), the global stream is the simpler approach. Some systems (such as Event Store) give you the best of both worlds, by having very fine-grained streams (such as per aggregate instance) but with the ability to combine them into larger streams (per stream type/category/partition, per multiple stream types, etc.) in a performant and predictable way out of the box, while still being simple by only requiring you to keep track of a single global event position.
If you go partitioned with Kafka:
Your projection code will need to handle concurrent consumer groups accessing the same read models when processing events for different partitions that need to go into the same models. Depending on your target store for the projection, there are lots of ways to handle this (transactions, optimistic concurrency, atomic operations, etc.) but it would be a problem for some target stores
Your projection code will need to keep track of the stream position of each partition, not just a single position. If your projection reads from multiple streams, it has to keep track of lots of positions.
Using a global stream removes both of those concerns - performance is usually likely to be good enough.
In either case, you'll likely also want to get the stream position into the long term event storage (i.e. Cassandra) - you could do this by having a dedicated process reading from the event stream (partitioned or global) and just updating the events in Cassandra with the global or partition position of each event. (I have a similar thing with MongoDB - I have a process reading the 'oplog' and copying oplog timestamps into events, since oplog timestamps are totally ordered).
Another option is to drop Cassandra from the initial command processing and use Kafka Streams instead:
Partitioned command stream is processed by joining with a partitioned KTable of aggregates
Command result and events are computed
Atomically, KTable is updated with changed aggregate, events are written to event stream and command response is written to command response stream.
You would then have a downstream event processor that copies the events into Cassandra for easier querying etc. (and which can add the Kafka stream position to each event as it does it to give the category ordering). This can help with catch up subscriptions, etc. if you don't want to use Kafka for long term event storage. (To catch up, you'd just read as far as you can from Cassandra and then switch to streaming from Kafka from the position of the last Cassandra event). On the other hand, Kafka itself can store events for ever, so this isn't always necessary.
I hope this helps a bit with understanding the tradeoffs and problems you might encounter.
Our team is trying to build a predictive maintenance system whose task is to look at a set of events and predict whether these events depict a set of known anomalies or not.
We are at the design phase and the current system design is as follows:
The events may occur on multiple sources of an IoT system (such as cloud platform, edge devices or any intermediate platforms)
The events are pushed by the data sources into a message queueing system (currently we have chosen Apache Kafka).
Each data source has its own queue (Kafka Topic).
From the queues, the data is consumed by multiple inference engines (which are actually neural networks).
Depending upon the feature set, an inference engine will subscribe to
multiple Kafka topics and stream data from those topics to continuously output the inference.
The overall architecture follows the single-responsibility principle meaning that every component will be separate from each other and run inside a separate Docker container.
Problem:
In order to classify a set of events as an anomaly, the events have to occur in the same time window. e.g. say there are three data sources pushing their respective events into Kafka topics, but due to some reason, the data is not synchronized.
So one of the inference engines pulls the latest entries from each of the kafka topics, but the corresponding events in the pulled data do not belong to the same time window (say 1 hour). That will result in invalid predictions due to out-of-sync data.
Question
We need to figure out how can we make sure that the data from all three sources are pushed in-order so that when an inference engine requests entries (say the last 100 entries) from multiple kakfa topics, the corresponding entries in each topic belong to the same time window?
I would suggest KSQL, which is a streaming SQL engine that enables real-time data processing against Apache Kafka. It also provides nice functionality for Windowed Aggregation etc.
There are 3 ways to define Windows in KSQL:
hopping windows, tumbling windows, and session windows. Hopping and
tumbling windows are time windows, because they're defined by fixed
durations they you specify. Session windows are dynamically sized
based on incoming data and defined by periods of activity separated by
gaps of inactivity.
In your context, you can use KSQL to query and aggregate the topics of interest using Windowed Joins. For example,
SELECT t1.id, ...
FROM topic_1 t1
INNER JOIN topic_2 t2
WITHIN 1 HOURS
ON t1.id = t2.id;
Some suggestions -
Handle delay at the producer end -
Ensure all three producers always send data in sync to Kafka topics by using batch.size and linger.ms.
eg. if linger.ms is set to 1000, all messages would be sent to Kafka within 1 second.
Handle delay at the consumer end -
Considering any streaming engine at the consumer side (be it Kafka-stream, spark-stream, Flink), provides windows functionality to join/aggregate stream data based on keys while considering delayed window function.
Check this - Flink windows for reference how to choose right window type link
To handle this scenario, data sources must provide some mechanism for the consumer to realize that all relevant data has arrived. The simplest solution is to publish a batch from data source with a batch Id (Guid) of some form. Consumers can then wait until the next batch id shows up marking the end of the previous batch. This approach assumes sources will not skip a batch, otherwise they will get permanently mis-aligned. There is no algorithm to detect this but you might have some fields in the data that show discontinuity and allow you to realign the data.
A weaker version of this approach is to either just wait x-seconds and assume all sources succeed in this much time or look at some form of time stamps (logical or wall clock) to detect that a source has moved on to the next time window implicitly showing completion of the last window.
The following recommendations should maximize success of event synchronization for the anomaly detection problem using timeseries data.
Use a network time synchronizer on all producer/consumer nodes
Use a heartbeat message from producers every x units of time with a fixed start time. For eg: the messages are sent every two minutes at the start of the minute.
Build predictors for producer message delay. use the heartbeat messages to compute this.
With these primitives, we should be able to align the timeseries events, accounting for time drifts due to network delays.
At the inference engine side, expand your windows at a per producer level to synch up events across producers.
I have time series data in Kafka. The schema is quite simple - the key is the channel name, and the values are Long/Double tuples of the timestamp and the value (in reality it's a custom Avro object but it boils down to this). They always come in correct chronological order.
The wanted end result is data packaged in 10 minute batches, aligned at 10 minutes (i.e., 00:00 < t <= 00:10, 00:10 < t <= 00:20, ..., 23: 50 < t <= 00:00). Each package is to contain only data of one channel.
My idea is to have two Spark Streaming jobs. The first one takes the data from the Kafka topics and dumps it to a table in a Cassandra database where the key is the timestamp and the channel name, and every time such an RDD hits a 10 minute boundary, this boundary is posted to another topic, alongside the channel whose boundary is hit.
The second job listens to this "boundary topic", and for every received 10 minute boundary, the data is pulled from Cassandra, some calculations like min, max, mean, stddev are done and the data and these results are packaged to a defined output directory. That way, each directory contains the data from one channel and one 10 minute window.
However, this looks a bit clunky and like a lot of extra work to me. Is this a feasible solution or are there any other more efficient tricks to it, like some custom windowing of the Kafka data?
I agree with your intuition that this solution is clunky. How about simply using the time windowing functionality built into the Streams DSL?
http://kafka.apache.org/11/documentation/streams/developer-guide/dsl-api.html#windowing
The most natural output would be a new topic containing the windowed aggregations, but if you really need it written to a directory that should be possible with Kafka Connect.
I work with the Flink Stream Processing, not Spark-streaming but I guess the programming concept of both of them is alike. So supposing data are ordered chronologically and you want to aggregate data for every 10 minutes and do some processing on aggregated data, I think the best approach is to use the Streaming Window Functions. I suggest to define a function to map every incoming data's timestamp to the last 10 minutes:
12:10:24 ----> 12:10:00
12:10:30 ----> 12:10:00
12:25:24 ----> 12:20:00
So you can create a keyed stream object like:
StreamObject<Long, Tuple<data>>
That the Long field is the mapped timestamp of every message. Then you can apply a window. You should search what kind of window is more appropriate for your case.
Point: Setting a key for the data stream will cause the window function to consider a logical window for every key.
In the most simple case, you should define a time window of 10 minutes and aggregate all data incoming on that period of time.
The other approach, if you know the rate of generating of data and how many messages will be generated in a period of 10 minutes, is to use Count window. For example, a window with the count of 20 will listen to the stream and aggregate all the messages with the same key in a logical window and apply the window function just when the number of messages in the window reaches 20.
After messages aggregated in a window as desired, you could apply your processing logic using a reduce function or some action like that.
We use "System Lag" to check the health of our Dataflow jobs. For example if we see an increase in system lag, we will try to see how to bring this metric down. There are few question regarding this metric.
1) What does system lag exactly means?
The maximum time that an item of data has been awaiting processing
Above is what we see in GCP Console when we hit information icon. What does an item of data mean in this case? Stream processing has concept of Windowing, event time vs processing time, watermark, etc. When is an item considered awaiting to be processed? For example is it simply when the message arrives regardless of its state?
2) What is the optimum threshold for this metric?
We try to keep this metric as low as possible, but we don't have any recommendation on how low we should keep it. For example do we have some recommendation such as keeping system lag between 20s to 30s is optimum.
3)How does system lag implicates sinks
How does system lag affect latency of the event itself?
Depending on the pipeline being executed there are a number of places that elements may be queued up awaiting processing. This is typically when the elements are passed between machines, such as within a GroupByKey, although the PubSub source also reflects the oldest unacked element.
For a given step (sinks included) "System Lag" measures the age of the oldest element in the closest input queue to that step.
It is not unusual for there to be spikes in this measure -- elements are pulled off the queue after they are processed, so if many new elements are delivered it may take a while before the queue is back to a manageable size. What is important is that the system lag goes back down after these spikes.
The latency of a sink depends on several factors:
The rate that elements arrive in the pipeline limits the rate the input watermark advances.
The configuration of windowing and triggers affect how long the pipeline must wait before emitting a given window.
System lag is a measure of how much added delay is currently being introduced by code executing within the pipeline.
It is likely easier to look at the "Data Watermark" of the sink, which reports up to what point in (event) time the sink has been processed.
I'm using DynamoDB Streams + Kinesis Client Library (KCL).
How can I measure latency between when an event was created in a stream and when it was processed on KCL side?
As I know, KCL's MillisBehindLatest metric is specific to Kinesis Streams(not DynamoDB streams).
approximateCreationDateTime record attribute has a minute-level approximation, which is not acceptable for monitoring in sub-second latency systems.
Could you please help with some useful metrics for monitoringDynamoDB Streams latency?
You can change the way you do writes in your application to allow your application to track the propagation delay of mutations in the table's stream. For example, you could always update a 'last_updated=' timestamp attribute when you create and update items. That way, when your creations and updates appear in the stream, you can estimate the propagation delay by subtracting the current time from last_updated in the NEW_IMAGE of the stream record.
Because deletions do not have a NEW_IMAGE in stream records, your deletes would need to take place in two steps:
logical deletion where you write the 'logically_deleted='
timestamp to the item and
physical deletion where you actually call DeleteItem immediately following 1.
Then, you would use the same math as for creations and updates, only differences being that you would use the OLD_IMAGE when processing deletions and you would need to subtract at least around 10ms to account for the time it takes to perform the logical delete (step 1).