Forwarded from Make. Build. Break. Reflect.
#aws #aurora #terraform #finops
Материал уровня senior
Мы умеем определять тип инстанса - по нагрузке на CPU/Memory и другим факторам.
Но насколько эффективно мы выбрали
Эффективно ли это спустя год или два?
А никто не знает, давайте разбираться.
- сперва пилим Dashboard к нашему существующему кластеру
https://gist.github.com/kruchkov-alexandr/d9335d7927e58d06557b994dc9f194de
- применяем, видим панель в CloudWatch
Сверху у нас панель чтения/записи хранилища, снизу размер базы данных (+снепшоты?).
Разделение нужно из-за разницы масштабов и удобства экспорта.
- выбираем период три месяца и кликаем на трех точках у обоих панелей и выбираем
- заходим в cost explorer и экспортируем дату за три месяца по кластеру
- заходим в вашу любимую AI(я спрашивал у клаудии, перплексити и грок3, все платные) и пилим промт нечто типа(можете писать свой, если мой вам кажется тупым):
"Help me decide if we should switch to Amazon Aurora I/O-Optimized. Use the attached billing screenshot/csv, three-month IOPS data from the CSV, and the IOPS/storage graphs to analyze our costs. Calculate our current I/O expenses, compare them to I/O-Optimized costs and check if our I/O costs exceed AWS’s 25% threshold for switching. Look at IOPS and storage trends, then recommend whether to switch, including specific cost figures. I’ve attached all files (billing, CSV, graphs).
based on this article
https://aws.amazon.com/blogs/database/estimate-cost-savings-for-the-amazon-aurora-i-o-optimized-feature-using-amazon-cloudwatch/"
- ждём ответа и все 4 нейронки мне выдали на 95% одинаковый подробнейший расчёт ответ. Вкратце "
- пишем менеджеру/боссу
I've analyzed our infrastructure costs over the last three months billing and IOPS data, to see if switching to Amazon Aurora I/O-Optimized makes sense. Right now, it's not cost-effective. Our I/O costs an average of $******* monthly (************ I/Os at $**** per million). Moving to I/O-Optimized would increase instance costs by ***%, from $******* to $******* - a $******* jump, which is $415.21 more than our current I/O expenses.
Our IOPS trends show peaks up to *** but no major growth, averaging ~** Write and ~**** Read IOPS hourly in February. Storage usage is growing at *** GB/month, but that doesn't impact the I/O-Optimized cost comparison. AWS suggests I/O-Optimized when I/O costs exceed **% of total Aurora spend, but ours ($******) are only **% of the $******* total, so we're below that threshold.
I recommend sticking with our standard configuration for now. We should keep monitoring I/O activity -if it exceeds **** I/Os monthly or I/O costs reach **% of our Aurora spend, we can revisit I/O-Optimized.
Прикладываем все файлы,скрины,расчёты.
- закрываем таску и трекаем время
Всё вышеописанное заняло у меня минут 15, а вот подготовительные работы(чтение про фичу, особенности, лимиты, как считать, написание борды, особенности биллинга и тп) почти половину дня.
* Если не верите ИИ, можете пересчитать вручную🐒
Дополнительные полезные ссылки(а вдруг вы мне не верите):
- анонс фичи
https://aws.amazon.com/about-aws/whats-new/2023/05/amazon-aurora-i-o-optimized/
- обзор менеджер уровня
https://aws.amazon.com/awstv/watch/b9bfc040ac5/
- пример расчётов (там руками считают, без ИИ)
https://aws.amazon.com/blogs/database/estimate-cost-savings-for-the-amazon-aurora-i-o-optimized-feature-using-amazon-cloudwatch/
Материал уровня senior
Мы умеем определять тип инстанса - по нагрузке на CPU/Memory и другим факторам.
Но насколько эффективно мы выбрали
Cluster storage configuration Авроры вашего проекта?Эффективно ли это спустя год или два?
А никто не знает, давайте разбираться.
- сперва пилим Dashboard к нашему существующему кластеру
https://gist.github.com/kruchkov-alexandr/d9335d7927e58d06557b994dc9f194de
- применяем, видим панель в CloudWatch
Сверху у нас панель чтения/записи хранилища, снизу размер базы данных (+снепшоты?).
Разделение нужно из-за разницы масштабов и удобства экспорта.
- выбираем период три месяца и кликаем на трех точках у обоих панелей и выбираем
Download as .csv и качаем оба файла- заходим в cost explorer и экспортируем дату за три месяца по кластеру
- заходим в вашу любимую AI(я спрашивал у клаудии, перплексити и грок3, все платные) и пилим промт нечто типа(можете писать свой, если мой вам кажется тупым):
"Help me decide if we should switch to Amazon Aurora I/O-Optimized. Use the attached billing screenshot/csv, three-month IOPS data from the CSV, and the IOPS/storage graphs to analyze our costs. Calculate our current I/O expenses, compare them to I/O-Optimized costs and check if our I/O costs exceed AWS’s 25% threshold for switching. Look at IOPS and storage trends, then recommend whether to switch, including specific cost figures. I’ve attached all files (billing, CSV, graphs).
based on this article
https://aws.amazon.com/blogs/database/estimate-cost-savings-for-the-amazon-aurora-i-o-optimized-feature-using-amazon-cloudwatch/"
- ждём ответа и все 4 нейронки мне выдали на 95% одинаковый подробнейший расчёт ответ. Вкратце "
Переходить пока рано".- пишем менеджеру/боссу
I've analyzed our infrastructure costs over the last three months billing and IOPS data, to see if switching to Amazon Aurora I/O-Optimized makes sense. Right now, it's not cost-effective. Our I/O costs an average of $******* monthly (************ I/Os at $**** per million). Moving to I/O-Optimized would increase instance costs by ***%, from $******* to $******* - a $******* jump, which is $415.21 more than our current I/O expenses.
Our IOPS trends show peaks up to *** but no major growth, averaging ~** Write and ~**** Read IOPS hourly in February. Storage usage is growing at *** GB/month, but that doesn't impact the I/O-Optimized cost comparison. AWS suggests I/O-Optimized when I/O costs exceed **% of total Aurora spend, but ours ($******) are only **% of the $******* total, so we're below that threshold.
I recommend sticking with our standard configuration for now. We should keep monitoring I/O activity -if it exceeds **** I/Os monthly or I/O costs reach **% of our Aurora spend, we can revisit I/O-Optimized.
Прикладываем все файлы,скрины,расчёты.
- закрываем таску и трекаем время
Всё вышеописанное заняло у меня минут 15, а вот подготовительные работы(чтение про фичу, особенности, лимиты, как считать, написание борды, особенности биллинга и тп) почти половину дня.
* Если не верите ИИ, можете пересчитать вручную
Дополнительные полезные ссылки(а вдруг вы мне не верите):
- анонс фичи
https://aws.amazon.com/about-aws/whats-new/2023/05/amazon-aurora-i-o-optimized/
- обзор менеджер уровня
https://aws.amazon.com/awstv/watch/b9bfc040ac5/
- пример расчётов (там руками считают, без ИИ)
https://aws.amazon.com/blogs/database/estimate-cost-savings-for-the-amazon-aurora-i-o-optimized-feature-using-amazon-cloudwatch/
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
🔥21👍3🤔2❤1👏1
MCP with Bedrock:
- Allows AI models to access information outside their built-in knowledge.
- Help build tools that enable AI models to perform actions (such as visiting websites or checking the weather).
- Establish communication through a standardized protocol between the user, the AI model and external tools.
https://community.aws/content/2uFvyCPQt7KcMxD9ldsJyjZM1Wp/model-context-protocol-mcp-and-amazon-bedrock
For example, if you ask "get me a summary of the blog post at this URL", the system will:
- Process your request
- Recognize it needs to use a tool to visit the webpage
- Fetch the content from the URL
- Return the information to the model
- Generate a summary based on the fetched content
#MCP #Bedrock
- Allows AI models to access information outside their built-in knowledge.
- Help build tools that enable AI models to perform actions (such as visiting websites or checking the weather).
- Establish communication through a standardized protocol between the user, the AI model and external tools.
https://community.aws/content/2uFvyCPQt7KcMxD9ldsJyjZM1Wp/model-context-protocol-mcp-and-amazon-bedrock
For example, if you ask "get me a summary of the blog post at this URL", the system will:
- Process your request
- Recognize it needs to use a tool to visit the webpage
- Fetch the content from the URL
- Return the information to the model
- Generate a summary based on the fetched content
#MCP #Bedrock
👍5🔥3
Всех с пятницей .
🔥 Выпущена версия 0.19.1 платформы для изучения SRE! 🔥
GitHub
📌 Что нового:
- обновлён docker image for runner(viktoruj/runner)
- добавлено api к ping-pong серверу
- добавлена эмуляция падения программы в ping-pong
- добавлена эмуляция медленного ответа в ping-pong
- добавлен просмотр текущих значений переменных в ping-pong
- добавлено изменение переменных в запущенном ping-pong через api
- добавлено получение текущих параметров системы через api в ping-pong
- добавлены скомпилированные bin ping-pong для различных операционных систем
документация ping-pong сервера
🧪 Доступные пробные экзамены:
CKA
CKAD
CKS
KCNA
KCSA
LFCS
Скрипты и видео с решениями экзаменов:
Да пребудет с вами сила!
🔥 Выпущена версия 0.19.1 платформы для изучения SRE! 🔥
GitHub
📌 Что нового:
- обновлён docker image for runner(viktoruj/runner)
- добавлено api к ping-pong серверу
- добавлена эмуляция падения программы в ping-pong
- добавлена эмуляция медленного ответа в ping-pong
- добавлен просмотр текущих значений переменных в ping-pong
- добавлено изменение переменных в запущенном ping-pong через api
- добавлено получение текущих параметров системы через api в ping-pong
- добавлены скомпилированные bin ping-pong для различных операционных систем
документация ping-pong сервера
🧪 Доступные пробные экзамены:
CKA
CKAD
CKS
KCNA
KCSA
LFCS
Скрипты и видео с решениями экзаменов:
Да пребудет с вами сила!
GitHub
GitHub - ViktorUJ/cks: Open-source Platform for learning kubernetes and aws eks and preparation for for Certified Kubernetes…
Open-source Platform for learning kubernetes and aws eks and preparation for for Certified Kubernetes exams (CKA ,CKS , CKAD) - GitHub - ViktorUJ/cks: Open-source Platform for learning kubern...
🔥33
Forwarded from valmont2k
дарю лайфхак для профи: на всех митигах можно тренировать слепую десятипальцевую печать, так даже полезнее.
всего несколько митингов и ты мастер
всего несколько митингов и ты мастер
👍13😎6
IngressNightmare — сразу несколько уязвимостей NGINX Controller for Kubernetes доступом к секретам всего и везде без авторизации:
https://www.wiz.io/blog/ingress-nginx-kubernetes-vulnerabilities
◽️ Кто пострадал — обладатели NGINX Controller версий до 1.12.1/1.11.5. Для устранения нужно срочно обновиться на последнюю версию.
◽️ Кто не пострадал — пользователи EKS:
EKS does not provide or install the ingress-nginx controller and is not affected by these issues.
Официальный отчёт о уязвимости Kubernetes:
https://kubernetes.io/blog/2025/03/24/ingress-nginx-cve-2025-1974/
#Kubernetes #security
https://www.wiz.io/blog/ingress-nginx-kubernetes-vulnerabilities
◽️ Кто пострадал — обладатели NGINX Controller версий до 1.12.1/1.11.5. Для устранения нужно срочно обновиться на последнюю версию.
◽️ Кто не пострадал — пользователи EKS:
EKS does not provide or install the ingress-nginx controller and is not affected by these issues.
Официальный отчёт о уязвимости Kubernetes:
https://kubernetes.io/blog/2025/03/24/ingress-nginx-cve-2025-1974/
#Kubernetes #security
wiz.io
CVE-2025-1974: The IngressNightmare in Kubernetes | Wiz Blog
Wiz Research uncovered RCE vulnerabilities (CVE-2025-1097, 1098, 24514, 1974) in Ingress NGINX for Kubernetes allowing cluster-wide secret access.
😁5👍2
We are excited to announce that on Friday, March 28, we are hosting an AWS Workshop: Amazon RDS for PostgreSQL, open to everyone!
With Amazon RDS, you can deploy scalable PostgreSQL instances in minutes with cost-efficient and resizable hardware capacity. Amazon RDS manages complex administrative tasks such as PostgreSQL software installation and upgrades, storage management, replication for high availability and read throughput, and backups for disaster recovery.
The workshop will be led by Sergey Pagin, an experienced AWS specialist and Lead Systems Engineer at EPAM Systems.
We will provide a dedicated training account for each participant who registers via the Google Form [https://forms.gle/L6Fp8md2XGwsK2DG8]. However, you can also use your personal AWS account.
⚠️ IMPORTANT: Please complete the registration form [https://forms.gle/L6Fp8md2XGwsK2DG8] by 21:00 UTC+1 on March 27.
🔗 Find out more details and register here:
https://wearecommunity.io/events/workshop-amazon-rds-for-postgresql
With Amazon RDS, you can deploy scalable PostgreSQL instances in minutes with cost-efficient and resizable hardware capacity. Amazon RDS manages complex administrative tasks such as PostgreSQL software installation and upgrades, storage management, replication for high availability and read throughput, and backups for disaster recovery.
The workshop will be led by Sergey Pagin, an experienced AWS specialist and Lead Systems Engineer at EPAM Systems.
We will provide a dedicated training account for each participant who registers via the Google Form [https://forms.gle/L6Fp8md2XGwsK2DG8]. However, you can also use your personal AWS account.
⚠️ IMPORTANT: Please complete the registration form [https://forms.gle/L6Fp8md2XGwsK2DG8] by 21:00 UTC+1 on March 27.
🔗 Find out more details and register here:
https://wearecommunity.io/events/workshop-amazon-rds-for-postgresql
👍9❤1💩1
Пишу тут про AI. Пишу в первую очередь для своих, во вторую для себя, чтобы в одном месте хранить материалы по теме.
Telegram
AIзбука
Ликбез (ликвидация безграмотности) по теме AI (Artificial Intelligence или ИИ — искусственный интеллект).
Предназначен для всех, но особенно для инженеров, потому что они особо тяжёлый случай.
Ссылка на канал: https://xn--r1a.website/AIzbuka
Контакт: @apple_rom
Предназначен для всех, но особенно для инженеров, потому что они особо тяжёлый случай.
Ссылка на канал: https://xn--r1a.website/AIzbuka
Контакт: @apple_rom
🔥8👍7
EKS Terraform demo
https://github.com/setheliot/eks_demo/
Deploys:
▪️ EKS cluster using EC2 nodes
▪️ DynamoDB table
▪️ EBS volume used as attached storage for the Kubernetes cluster (a PersistentVolume)
▪️ Demo "guestbook" application, deployed via containers
▪️ ALB to access the app
#EKS #Terraform
https://github.com/setheliot/eks_demo/
Deploys:
▪️ EKS cluster using EC2 nodes
▪️ DynamoDB table
▪️ EBS volume used as attached storage for the Kubernetes cluster (a PersistentVolume)
▪️ Demo "guestbook" application, deployed via containers
▪️ ALB to access the app
#EKS #Terraform
👍5💊5
Forwarded from AWS Weekly (Max Skutin)
▪️ Amplify Hosting WAF Protection | GA
▪️ Amplify
▫️ samples to Deploy Storage Browser for S3
▫️ Shared Keychain support for Swift
▪️ Application Recovery Controller FIS recovery action for zonal autoshift
▪️ Bedrock Custom Model Import introduces real-time cost transparency
▪️ Bedrock Guardrails industry-leading image content filters | GA
▪️ Bedrock Knowledge Bases Opensearch Managed Cluster for vector storage
▪️ CloudFormation targeted resource scans in the IaC generator
▪️ CodeBuild custom cache keys for S3 caching
▪️ Connected Mobility Solution new features
▪️ Database Insights customization of its metrics dashboard
▪️ DataZone metadata rules for publishing
▪️ Dedicated Local Zones gp3 and io1 ebs volumes
▪️ DMS Schema Conversion IBM Db2 for z/OS to RDS for Db2 conversion
▪️ DynamoDB percentile statistics for request latency
▪️ DynamoDB Streams PrivateLink support
▪️ EC2 more bandwidth and jumbo frames to select destinations
▪️ EKS enforces upgrade insights checks as part of cluster upgrades
▪️ Elemental MediaConnect NDI® outputs
▪️ EventBridge Scheduler PrivateLink support
▪️ GameLift Servers next-gen EC2 instances
▪️ IAM dual-stack (IPv4 and IPv6) environments
▪️ Keyspaces Multi-Region support for User Defined Types (UDTs)
▪️ Lambda Ruby 3.4
▪️ Marketplace new seller experiences for ML products
▪️ Network Firewall pass action rule alerts and JA4 filtering
▪️ Network Manager support PrivateLink and IPv6
▪️ Open Source Corretto 24 | GA
▪️ Parallel Computing Service Terraform support
▪️ Polly New Korean voice
▪️ Q Business upgrades for Slack and Teams Integrations
▪️ Q in QuickSight Scenarios capability | GA
▪️ RDS for MySQL Innovation Release 9.2 in Preview Environment
▪️ RDS for SQL Server linked servers to Teradata databases
▪️ Route 53 Profiles IPv6 Service Endpoints
▪️ SageMaker HyperPod multi-head node support in Slurm
▪️ SageMaker metadata rules to enforce standards and improve data governance
Please open Telegram to view this post
VIEW IN TELEGRAM
🔥2
Amazon Nova Act - new AI model trained to perform actions within a web browser:
https://labs.amazon.science/blog/nova-act
Amazon Nova Act SDK research preview is available from today:
https://nova.amazon.com
#Nova
https://labs.amazon.science/blog/nova-act
Amazon Nova Act SDK research preview is available from today:
https://nova.amazon.com
#Nova
Amazon
Amazon Nova
Try new AI capabilities with Amazon Nova foundation models. Build agents with the new agentic model Amazon Nova Act. Create and edit images with Amazon Nova Canvas. Engage in human-like voice conversations with Amazon Nova Sonic. Generate text and code, and…
🔥5👍4
EKS community Add-ons
https://aws.amazon.com/blogs/containers/announcing-amazon-eks-community-add-ons-catalog/
Теперь можно уменьшить набор стандартных чартов для установки и поставить их через эдд-оны:
metrics-server
kube-state-metrics
prometheus-node-exporter
cert-manager
external-dns
#EKS
https://aws.amazon.com/blogs/containers/announcing-amazon-eks-community-add-ons-catalog/
Теперь можно уменьшить набор стандартных чартов для установки и поставить их через эдд-оны:
metrics-server
kube-state-metrics
prometheus-node-exporter
cert-manager
external-dns
#EKS
🔥22👍12💩3❤2
🚀 Join our AWS Workshop: Amazon ECS Finhack on April 4! 🚀
The Amazon ECS Finhack workshop introduces a series of hands-on labs designed to help you understand key cost optimization levers for ECS container infrastructure.
The modules will cover:
✅ Cost Visibility – Understanding costs with container infrastructure
✅ Auto Scaling – Improving compute utilization with elasticity
✅ Compute Optimization – Optimizing compute costs with Spot and AWS Graviton
The workshop will be led by Aliaksandr Chernak, an experienced AWS specialist and Lead Systems Engineer at EPAM Systems.
We will provide a dedicated training account for each participant who registers via the Google Form [https://forms.gle/9RdYT9kmh6hzwMVn7]. However, you can also use your personal AWS account.
⚠️ IMPORTANT: Please complete the registration form [https://forms.gle/9RdYT9kmh6hzwMVn7] by 21:00 CET on April 3.
🔗 Find out more details and register here:
https://wearecommunity.io/events/amazon-ecs-finhack-workshop
The Amazon ECS Finhack workshop introduces a series of hands-on labs designed to help you understand key cost optimization levers for ECS container infrastructure.
The modules will cover:
✅ Cost Visibility – Understanding costs with container infrastructure
✅ Auto Scaling – Improving compute utilization with elasticity
✅ Compute Optimization – Optimizing compute costs with Spot and AWS Graviton
The workshop will be led by Aliaksandr Chernak, an experienced AWS specialist and Lead Systems Engineer at EPAM Systems.
We will provide a dedicated training account for each participant who registers via the Google Form [https://forms.gle/9RdYT9kmh6hzwMVn7]. However, you can also use your personal AWS account.
⚠️ IMPORTANT: Please complete the registration form [https://forms.gle/9RdYT9kmh6hzwMVn7] by 21:00 CET on April 3.
🔗 Find out more details and register here:
https://wearecommunity.io/events/amazon-ecs-finhack-workshop
❤2👍2👀2
KAI (Kubernetes AI) Scheduler from NVidia - dynamically allocate GPU resources to workloads:
https://github.com/NVIDIA/KAI-Scheduler
#Kubernetes
https://github.com/NVIDIA/KAI-Scheduler
#Kubernetes
GitHub
GitHub - NVIDIA/KAI-Scheduler: KAI Scheduler is an open source Kubernetes Native scheduler for AI workloads at large scale
KAI Scheduler is an open source Kubernetes Native scheduler for AI workloads at large scale - NVIDIA/KAI-Scheduler
👍11❤5🎉1
Forwarded from Christine Armiyants
Here are some ways you can work with MCP across AWS today:
🟠 Bedrock Agents now supports MCP tools through the new Inline Agents API ➡️ here
🟠 Learn how to run MCP servers and clients on AWS using open source MCP libraries ➡️ here
🟠 Get started with open-source AWS MCP servers for code assistants ➡️ here
More integrations are coming soon☺️
🟠 Bedrock Agents now supports MCP tools through the new Inline Agents API ➡️ here
🟠 Learn how to run MCP servers and clients on AWS using open source MCP libraries ➡️ here
🟠 Get started with open-source AWS MCP servers for code assistants ➡️ here
More integrations are coming soon☺️
🥱5👍1
VPC + BGP = VPC Route Server with dynamic routing inside VPC
https://docs.aws.amazon.com/vpc/latest/userguide/dynamic-routing-route-server.html
You can now dynamically update VPC and internet gateway route tables with your IPv4 or IPv6 routes.
Features include:
- Automatically update route tables with routes learned from BGP peers
- Propagation of the best available routes based on BGP attributes
- Dynamically update routes when network conditions change
- Failure endpoints using BFD
There are still some gaps like the support for route tables associated with virtual private gateways or propagation of routes into a transit gateway route table.
#VPC
https://docs.aws.amazon.com/vpc/latest/userguide/dynamic-routing-route-server.html
You can now dynamically update VPC and internet gateway route tables with your IPv4 or IPv6 routes.
Features include:
- Automatically update route tables with routes learned from BGP peers
- Propagation of the best available routes based on BGP attributes
- Dynamically update routes when network conditions change
- Failure endpoints using BFD
There are still some gaps like the support for route tables associated with virtual private gateways or propagation of routes into a transit gateway route table.
#VPC
👍15😱9🔥3
Intel vs AMD vs Graviton: Performance & Price
https://www.youtube.com/watch?v=kV6YizoFuPU
TL;DR:
▫️ Graviton is the best
▫️ AMD is the fastest
▫️ Intel is Intel
#cost_optimization
https://www.youtube.com/watch?v=kV6YizoFuPU
TL;DR:
▫️ Graviton is the best
▫️ AMD is the fastest
▫️ Intel is Intel
#cost_optimization
YouTube
Intel vs AMD vs Graviton: Performance & Price
👨💼📈 Mentorship/On-the-Job Support/Consulting - https://calendly.com/antonputra/youtube or me@antonputra.com
=========
🤝 - LinkedIn - https://www.linkedin.com/in/anton-putra
🎙 - Twitter - https://twitter.com/antonvputra
📧 - Email - me@antonputra.com
👨💻…
=========
🤝 - LinkedIn - https://www.linkedin.com/in/anton-putra
🎙 - Twitter - https://twitter.com/antonvputra
📧 - Email - me@antonputra.com
👨💻…
😁23👍5🤮2🔥1
Начиная с Terraform AWS Provider v6.0.0 алиасы уже не нужны — можно указывать AWS Region непосредственно в ресурсах, используя лишь один провайдер.
#Terraform
#Terraform
👏42👍9🔥9❤6😍1👀1🦄1