For the past 1 month, I had a chance to work with AWS Lambda. During the period of work with Lambda, I collected a lot of thoughts about this technology and would like to share them with you. Getting started So if you don’t know anything about AWS, I recommend starting with official docs: Amazon has a very rich documentation which will explain all the details about Lambda. If you don’t want to read the whole doc, then Lambda is a technology which allows you to deploy your code in a so-called Lambda functions - a containers somewhere inside AWS infrastructure. This gives a lot of benefits: you pay money only when you start invoking Lambda. The pricing for it is relatively low, as usual, AWS has a free tier which includes 1M free requests per month and 400,000 GB-seconds of compute time per month. The free tier description is a big confusing, I recommend using this table: Memory (MB) Free tier seconds per month Price per 100ms ($) 128 3,200,000 0.000000208 … … … 512 800,000 0.000000834 … … … 1024 400,000 0.000001667 … … … 2048 200,000 0.000003334 … … … 3008 136,170 0.000004897 Basically, for each particular...
Today is the starting of the fourth year since I began my journey with microservices. I started with a theoretical knowledge about this architecture and now I ended up with a more deep and practical experience. While I still believe I can find news problems in microservices, I prepared an article with a list of problems which I had a chance to face in my work. You will read short stories which I faced during my work. If you don’t agree with them and think that they could be fixed and identified earlier - that’s okay, I believe that you can’t find a microservices structure with identical problems - every organization has its own path and its own problems, thus, things that failed in one microservices architecture, could be omitted in another. Isolated messaging layer This story is about the messaging layer. You know the story, every cool microservices architecture has to have its own messaging layer: the idea is that you have an asynchronous way of communication between your services. I spent some time explaining this in Microservices interaction at scale using Apache Kafka article. So, the perfect scenario assumes that you have a bunch of microservices and some...
I’m working with Kafka for more than 2 years and I wasn’t sure if Kafka Consumer eats more RAM memory when it has more partitions. I couldn’t find any useful information on the internet, so I decided to measure everything by myself. Inputs I started with 1 broker, since I am interested in actual memory consumption for 1 and 1000 partition topics. I know, lauching Kafka in a cluster can differs, because we have replication processes, acknowledgments, and other cluster things, but let’s skip it for now. Two basic commands for launching Kafka single node cluster: bin/zookeeper-server-start.sh config/zookeeper.properties bin/kafka-server-start.sh config/server.properties I created two topics, topic1, with 1 partition, and topic2, with 1000 partitions. I believe, the difference between partitions is enough for understanding memory consumption. bin/kafka-topics.sh --create --zookeeper localhost:2181 --replication-factor 1 --partitions 1 --topic topic1 bin/kafka-topics.sh --create --zookeeper localhost:2181 --replication-factor 1 --partitions 1000 --topic topic2 It’s good that Kafka provides us with kafka-producer-perf-test.sh, a performance script, which let us load test Kafka. bin/kafka-producer-perf-test.sh --topic topic1 --num-records 99999999999999 --throughput 1 --producer-props bootstrap.servers=localhost:9092 key.serializer=org.apache.kafka.common.serialization.StringSerializer value.serializer=org.apache.kafka.common.serialization.StringSerializer --record-size 100 So, I consequently launched load tests to insert data into two topics with a throughput of 1, 200, 500 and 1000 messages/second. I collected all...
Hi, my name is Ivan Ursul and I am a freelance engineer since 2015. It’s been a while since I started my career as an independent freelancer. I started it as an engineer in Upwork, in one of their teams, where I was involved first in reporting backend service, then in the time-tracker pipeline, which is served as a backend for Upwork Tracker Application(UTA) client. Today I continue my work with Upwork, but I am also actively working with other customers, who are very different. I’ve successfully completed 22 projects since the very beginning. That’s why I decided to write an article about different aspects of everyday life of a freelance engineer. You may agree or disagree, anyway, I encourage you to leave your thoughts under this article. This article will be grounded on Upwork platform, I haven’t used other platforms, but I am quite sure the approach is the same. Learn your customer You will have to find out the common things about your clients. Are all of them technical? Do you prefer to work with non-technical people? What is your industry domain? These are the questions you should have answers to. After you realize what combines your customers...
Few days ago I had a problem on one of the projects that I am working on, we had a memory leak. During the two days period our services crashed three times, so I decided to investigate it. Everything which I’m going to talk about is not a rocket science, there’s no clever and tricky tips, it’s just a straighforward explanation how you can find memory leaks. Exposing JMX I had a problem on a production instance, so I started my services with JMX feature enabled. Just start your apps with following params: -Djavax.management.builder.initial= -Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=${whatever_port} -Dcom.sun.management.jmxremote.authenticate=false -Dcom.sun.management.jmxremote.ssl=false Starting JVisualVM Just enter your terminal and type jvisualvm. You should get following screen: Add a remote connection, specify JMX port and connect. Waiting You have to wait some time before retained memory will take place and you will be able to analyse it. It’s up to you how long to wait, in my case, it was enough to wait 4-5 hours to get 100% proof of what part of the system is leaking. Getting heap dump Now go to Monitor section, press Heap Dump button and specify path where heap dump should be saved. In my case it was /tmp/**.hproof. Then...
