I have a Kafka broker running in k8s and I notice that I have a regular problem when we write in disk.
I have two metrics, the income bytes in the broker, and the bytes written in disk.
As you can see in the graph the income is quite stable, but when you see the graph below, you can see the write in disk is more unstable, and sometimes go slower than 30mb/s when the income mb/s never go below 80 mb/s
Looking the resources of the Kafka broker there are enough memory and cpu.
Is this a problem of the disk?, running dd command we can write 500 mb/s
Here I create another panel where you can see how 4 times we are around 1 minute writing in disk a lot less that the income bytes. Those gaps is killing my app with OOM
Related
I'm trying to load messages from relatively large topic (billion+ records, more then 100 GiB, single partition) using Apache NiFi (nifi-1.11.4-RC1, OpenJDK 8, RHEL7), but performance seems to be far too low:
1248429 messages (276.2 MB) per 5 minutes for ConsumeKafka_2_0 and 295 batches (282.5 MB) for ConsumeKafkaRecord_2_0. I.e. only 4161 messages (920 KB) per second.
Results of kafka-consumer-perf-test.sh (same node, same consumer group and same topic) are more impressive:
263.4 MB (1190937 records) per second. Too much difference for any reasonable overhead.
I've configured cluster according to Best practices for setting up a high performance NiFi installation, but throughput didn't increase.
Each node has 256 GB RAM and 20 cores, Maximum Timer Driven Thread Count is set to 120, but NiFi GUI shows only 1 or 2 active threads, and CPU load is almost zero, so is disk queue.
I've tested several flows, but even ConsumeKafka_2_0 with autoterminated 'success' relationship shows the same speed.
Is it possible to increase performance of these processors? It looks like some artificial limit or throttle, because I couldn't find any bottleneck...
Help, please, I'm completely stuck!
UPD1:
# JVM memory settings
java.arg.2=-Xms10240m
java.arg.3=-Xmx10240m
Scheduling Strategy : Timer driven
Concurrent Tasks : 64
Run Schedule : 0 sec
Execution : All nodes
Maximum Timer Driven Thread Count : 120
Maximum Event Driven Thread Count : 20
UPD2:
When I consume topic with many partitions or several topics together with one ConsumeKafka_2_0 processor, or when I use several processors with different consumer groups with same topic, total throughput increases accordingly.
So, Maximum Timer Driven Thread Count and Concurrent Tasks aren't primary culprits. Problem is somewhere in task scheduling, or in processor itself.
We've had success increasing ConsumeKafka throughput by changing the processor's yield duration from 1 to 0 seconds and increasing the socket's buffer size to 1 MB.
receive.buffer.bytes=1048576
You may find other things to try here:
https://blog.newrelic.com/engineering/kafka-best-practices/
I am building a production environment where I will be having Apache Kafka. I want to know the best hardware combination to have for better performance. I will be having 5000 transactions/second.
You would need to provide some more details regarding your use-case like average size of messages etc. but here's my 2 cents anyway:
Confluent's documentation might shed some light:
CPUs Most Kafka deployments tend to be rather light on CPU
requirements. As such, the exact processor setup matters less than the
other resources. Note that if SSL is enabled, the CPU requirements can
be significantly higher (the exact details depend on the CPU type and
JVM implementation).
You should choose a modern processor with multiple cores. Common
clusters utilize 24 core machines.
If you need to choose between faster CPUs or more cores, choose more
cores. The extra concurrency that multiple cores offers will far
outweigh a slightly faster clock speed.
How to compute your throughput
It might also be helpful to compute the throughput. For example, if you have 800 messages per second, of 500 bytes each then your throughput is 800*500/(1024*1024) = ~0.4MB/s. Now if your topic is partitioned and you have 3 brokers up and running with 3 replicas that would lead to 0.4/3*3=0.4MB/s per broker.
More details regarding your architecture can be found in Confluent's whitepaper Apache Kafka and Confluent Reference Architecture. Here's the section for memory usage,
ZooKeeper uses the JVM heap, and 4GB RAM is typically sufficient. Too
small of a heap will result in high CPU due to constant garbage
collection while too large heap may result in long garbage collection
pauses and loss of connectivity within the ZooKeeper cluster.
Kafka brokers use both the JVM heap and the OS page cache. The JVM heap is used for replication of partitions between brokers and for log
compaction. Replication requires 1MB (default replica.max.fetch.size)
for each partition on the broker. In Apache Kafka 0.10.1 (Confluent
Platform 3.1), we added a new configuration
(replica.fetch.response.max.bytes) that limits the total RAM used for
replication to 10MB, to avoid memory and garbage collection issues
when the number of partitions on a broker is high. For log compaction,
calculating the required memory is more complicated and we recommend
referring to the Kafka documentation if you are using this feature.
For small to medium-sized deployments, 4GB heap size is usually
sufficient. In addition, it is highly recommended that consumers
always read from memory, i.e. from data that was written to Kafka and
is still stored in the OS page cache. The amount of memory this
requires depends on the rate at this data is written and how far
behind you expect consumers to get. If you write 20GB per hour per
broker and you allow brokers to fall 3 hours behind in normal
scenario, you will want to reserve 60GB to the OS page cache. In cases
where consumers are forced to read from disk, performance will drop
significantly
Kafka Connect itself does not use much memory, but some connectors buffer data internally for efficiency. If you run multiple connectors
that use buffering, you will want to increase the JVM heap size to 1GB
or higher.
Consumers use at least 2MB per consumer and up to 64MB in cases of large responses from brokers (typical for bursty traffic).
Producers will have a buffer of 64MB each. Start by allocating 1GB RAM and add 64MB for each producer and 16MB for each consumer planned.
There are many different factors that need to be taken into consideration when it comes to tune the configuration of your architecture. I would suggest to go through the aforementioned documentation, monitor your existing cluster and resources and finally tune them accordingly.
How can i calculate how much memory and cpu my Kafka cluster needs?
My cluster consists from 3 nodes, with throughput of ~800 messages per second.
Currently they have (each) 6 GB ram, 2 CPU, 1T disk, and it seems to be not enough. How much would you allocate?
You would need to provide some more details regarding your use-case like average size of messages etc. but here's my 2 cents anyway:
Confluent's documentation might shed some light:
CPUs Most Kafka deployments tend to be rather light on CPU
requirements. As such, the exact processor setup matters less than the
other resources. Note that if SSL is enabled, the CPU requirements can
be significantly higher (the exact details depend on the CPU type and
JVM implementation).
You should choose a modern processor with multiple cores. Common
clusters utilize 24 core machines.
If you need to choose between faster CPUs or more cores, choose more
cores. The extra concurrency that multiple cores offers will far
outweigh a slightly faster clock speed.
How to compute your throughput
It might also be helpful to compute the throughput. For example, if you have 800 messages per second, of 500 bytes each then your throughput is 800*500/(1024*1024) = ~0.4MB/s. Now if your topic is partitioned and you have 3 brokers up and running with 3 replicas that would lead to 0.4/3*3=0.4MB/s per broker.
More details regarding your architecture can be found in Confluent's whitepaper Apache Kafka and Confluent Reference Architecture. Here's the section for memory usage,
ZooKeeper uses the JVM heap, and 4GB RAM is typically sufficient. Too
small of a heap will result in high CPU due to constant garbage
collection while too large heap may result in long garbage collection
pauses and loss of connectivity within the ZooKeeper cluster.
Kafka brokers use both the JVM heap and the OS page cache. The JVM heap is used for replication of partitions between brokers and for log
compaction. Replication requires 1MB (default replica.max.fetch.size)
for each partition on the broker. In Apache Kafka 0.10.1 (Confluent
Platform 3.1), we added a new configuration
(replica.fetch.response.max.bytes) that limits the total RAM used for
replication to 10MB, to avoid memory and garbage collection issues
when the number of partitions on a broker is high. For log compaction,
calculating the required memory is more complicated and we recommend
referring to the Kafka documentation if you are using this feature.
For small to medium-sized deployments, 4GB heap size is usually
sufficient. In addition, it is highly recommended that consumers
always read from memory, i.e. from data that was written to Kafka and
is still stored in the OS page cache. The amount of memory this
requires depends on the rate at this data is written and how far
behind you expect consumers to get. If you write 20GB per hour per
broker and you allow brokers to fall 3 hours behind in normal
scenario, you will want to reserve 60GB to the OS page cache. In cases
where consumers are forced to read from disk, performance will drop
significantly
Kafka Connect itself does not use much memory, but some connectors buffer data internally for efficiency. If you run multiple connectors
that use buffering, you will want to increase the JVM heap size to 1GB
or higher.
Consumers use at least 2MB per consumer and up to 64MB in cases of large responses from brokers (typical for bursty traffic).
Producers will have a buffer of 64MB each. Start by allocating 1GB RAM and add 64MB for each producer and 16MB for each consumer planned.
There are many different factors that need to be taken into consideration when it comes to tune the configuration of your architecture. I would suggest to go through the aforementioned documentation, monitor your existing cluster and resources and finally tune them accordingly.
I think you want to start by profiling your kafka cluster.
See the answer to this post: CPU Profiling kafka brokers.
It basically recommends that you use a prometheus and grafana stack to visualize your load on a timeline - from this you should be able to determine your bottleneck. And links to an article that describes how.
Also, you may find the post interresting, because the poster seems to have about the same workload as you.
Below is the configuration:
2 JBoss application nodes
5 listeners on the application node with 50 threads each, supports
clustering and is set up as active-active listener, so they run on
both app nodes
The listener simply gets the message and logs the information into
database
50000 messages are posted into ActiveMQ using JMeter.
Here is the observation on first execution:
Total 50000 messages are consumed in approx 22 mins.
first 0-10000 messages consumed in 1 min approx
10000-20000 messages consumed in 2 mins approx
20000-30000 messages consumed in 4 mins approx
30000-40000 messages consumed in 6 mins approx
40000-50000 messages consumed in 8 mins
So we see the message consumption time is increasing with increasing number of messages.
Second execution without restarting any of the servers:
50000 messages consumed in 53 mins approx!
But after deleting data folder of activemq and restarting activemq,
performance again improves but degrades as more data enters the queue!
I tried multiple configuration in activemq.xml, but no success...
Anybody faced similar issue, and got any solution ? Let me know. Thanks.
I've seen similar slowdowns in our production systems when pending message counts go high. If you're flooding the queues then the MQ process can't keep all the pending messages in memory, and has to go to disk to serve a message. Performance can fall off a cliff in these circumstances. Increase the memory given to the MQ server process.
Also looks as though the disk storage layout is not particularly efficient - perhaps having each message as a file in a single directory? This can make access time rise as traversing disk directory takes longer.
50000 messages in > 20 mins seems very low performance.
Following configuration works well for me (these are just pointers. You may already have tried some of these but see if it works for you)
1) Server and queue/topic policy entry
// server
server.setDedicatedTaskRunner(false)
// queue policy entry
policyEntry.setMemoryLimit(queueMemoryLimit); // 32mb
policyEntry.setOptimizedDispatch(true);
policyEntry.setLazyDispatch(true);
policyEntry.setReduceMemoryFootprint(true);
policyEntry.setProducerFlowControl(true);
policyEntry.setPendingQueuePolicy(new StorePendingQueueMessageStoragePolicy());
2) If you are using KahaDB for persistence then use per destination adapter (MultiKahaDBPersistenceAdapter). This keeps the storage folders separate for each destination and reduces synchronization efforts. Also if you do not worry about abrupt server restarts (due to any technical reason) then you can reduce then disk sync efforts by
kahaDBPersistenceAdapter.setEnableJournalDiskSyncs(false);
3) Try increasing the memory usage, temp and storage disk usage values at server level.
4) If possible increase prefetchSize in prefetch policy. This will improve performance but also increases the memory footprint of consumers.
5) If possible use transactions in consumers. This will help to reduce the message acknowledgement handling and disk sync efforts by server.
Point 5 mentioned by #hemant1900 solved the problem :) Thanks.
5) If possible use transactions in consumers. This will help to reduce
the message acknowledgement handling and disk sync efforts by server.
The problem was in my code. I had not used transaction to persist the data in consumer, which is anyway bad programming..I know :(
But didn't expect that could have caused this issue.
Now 50000, messages are getting processed in less than 2 mins.
Scenario: I have a low-volume topic (~150msgs/sec) for which we would like to have a
low propagation delay from producer to consumer.
I added a time stamp from a producer and read it at consumer to record the propagation delay, with default configurations the msg (of 20 bytes) showed a propagation delay of 1960ms to 1230ms. No network delay is involved since, I tried on a 1 producer and 1 simple consumer on the same machine.
When I have tried adjusting the topic flush interval to 20ms, it drops
to 1100ms to 980ms. Then I tried adjusting the consumers "fetcher.backoff.ms" to 10ms, it dropped to 1070ms - 860ms.
Issue: For a 20 bytes of a msg, I would like to have a propagation delay as low as possible and ~950ms is a higher figure.
Question: Anything I am missing out in configuration?
I do welcome comments, delay which you got as minimum.
Assumption: The Kafka system involves the disk I/O before the consumer get the msg from the producer and this goes with the hard disk RPM and so on..
Update:
Tried to tune the Log Flush Policy for Durability & Latency.Following is the configuration:
# The number of messages to accept before forcing a flush of data to disk
log.flush.interval=10
# The maximum amount of time a message can sit in a log before we force a flush
log.default.flush.interval.ms=100
# The interval (in ms) at which logs are checked to see if they need to be
# flushed to disk.
log.default.flush.scheduler.interval.ms=100
For the same msg of 20 bytes, the delay was 740ms -880ms.
The following statements are made clear in the configuration itself.
There are a few important trade-offs:
Durability: Unflushed data is at greater risk of loss in the event of a crash.
Latency: Data is not made available to consumers until it is flushed (which adds latency).
Throughput: The flush is generally the most expensive operation.
So, I believe there is no way to come down to a mark of 150ms - 250ms. (without hardware upgrade) .
I am not trying to dodge the question but I think that kafka is a poor choice for this use case. While I think Kafka is great (I have been a huge proponent of its use at my workplace), its strength is not low-latency. Its strengths are high producer throughput and support for both fast and slow consumers. While it does provide durability and fault tolerance, so do more general purpose systems like rabbitMQ. RabbitMQ also supports a variety of different clients including node.js. Where rabbitMQ falls short when compared to Kafka is when you are dealing with extremely high volumes (say 150K msg/s). At that point, Rabbit's approach to durability starts to fall apart and Kafka really stands out. The durability and fault tolerance capabilities of rabbit are more than capable at 20K msg/s (in my experience).
Also, to achieve such high throughput, Kafka deals with messages in batches. While the batches are small and their size is configurable, you can't make them too small without incurring a lot of overhead. Unfortunately, message batching makes low-latency very difficult. While you can tune various settings in Kafka, I wouldn't use Kafka for anything where latency needed to be consistently less than 1-2 seconds.
Also, Kafka 0.7.2 is not a good choice if you are launching a new application. All of the focus is on 0.8 now so you will be on your own if you run into problems and I definitely wouldn't expect any new features. For future stable releases, follow the link here stable Kafka release
Again, I think Kafka is great for some very specific, though popular, use cases. At my workplace we use both Rabbit and Kafka. While that may seem gratuitous, they really are complimentary.
I know it's been over a year since this question was asked, but I've just built up a Kafka cluster for dev purposes, and we're seeing <1ms latency from producer to consumer. My cluster consists of three VM nodes running on a cloud VM service (Skytap) with SAN storage, so it's far from ideal hardware. I'm using Kafka 0.9.0.0, which is new enough that I'm confident the asker was using something older. I have no experience with older versions, so you might get this performance increase simply from an upgrade.
I'm measuring latency by running a Java producer and consumer I wrote. Both run on the same machine, on a fourth VM in the same Skytap environment (to minimize network latency). The producer records the current time (System.nanoTime()), uses that value as the payload in an Avro message, and sends (acks=1). The consumer is configured to poll continuously with a 1ms timeout. When it receives a batch of messages, it records the current time (System.nanoTime() again), then subtracts the receive time from the send time to compute latency. When it has 100 messages, it computes the average of all 100 latencies and prints to stdout. Note that it's important to run the producer and consumer on the same machine so that there is no clock sync issue with the latency computation.
I've played quite a bit with the volume of messages generated by the producer. There is definitely a point where there are too many and latency starts to increase, but it's substantially higher than 150/sec. The occasional message takes as much as 20ms to deliver, but the vast majority are between 0.5ms and 1.5ms.
All of this was accomplished with Kafka 0.9's default configurations. I didn't have to do any tweaking. I used batch-size=1 for my initial tests, but I found later that it had no effect at low volume and imposed a significant limit on the peak volume before latencies started to increase.
It's important to note that when I run my producer and consumer on my local machine, the exact same setup reports message latencies in the 100ms range -- the exact same latencies reported if I simply ping my Kafka brokers.
I'll edit this message later with sample code from my producer and consumer along with other details, but I wanted to post something before I forget.
EDIT, four years later:
I just got an upvote on this, which led me to come back and re-read. Unfortunately (but actually fortunately), I no longer work for that company, and no longer have access to the code I promised I'd share. Kafka has also matured several versions since 0.9.
Another thing I've learned in the ensuing time is that Kafka latencies increase when there is not much traffic. This is due to the way the clients use batching and threading to aggregate messages. It's very fast when you have a continuous stream of messages, but any time there is a moment of "silence", the next message will have to pay the cost to get the stream moving again.
It's been some years since I was deep in Kafka tuning. Looking at the latest version (2.5 -- producer configuration docs here), I can see that they've decreased linger.ms (the amount of time a producer will wait before sending a message, in hopes of batching up more than just the one) to zero by default, meaning that the aforementioned cost to get moving again should not be a thing. As I recall, in 0.9 it did not default to zero, and there was some tradeoff to setting it to such a low value. I'd presume that the producer code has been modified to eliminate or at least minimize that tradeoff.
Modern versions of Kafka seem to have pretty minimal latency as the results from here show:
2 ms (median)
3 ms (99th percentile)
14 ms (99.9th percentile)
Kafka can achieve around millisecond latency, by using synchronous messaging. With synchronous messaging, the producer does not collect messages into a patch before sending.
bin/kafka-console-producer.sh --broker-list my_broker_host:9092 --topic test --sync
The following has the same effect:
--batch-size 1
If you are using librdkafka as Kafka client library, you must also set socket.nagle.disable=True
See https://aivarsk.com/2021/11/01/low-latency-kafka-producers/ for some ideas on how to see what is taking those milliseconds.