<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Ravi Yasakeerthi on Medium]]></title>
        <description><![CDATA[Stories by Ravi Yasakeerthi on Medium]]></description>
        <link>https://medium.com/@lkravi?source=rss-4844d78af5------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*EShI5jiHhMPGI22D</url>
            <title>Stories by Ravi Yasakeerthi on Medium</title>
            <link>https://medium.com/@lkravi?source=rss-4844d78af5------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 19:32:35 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@lkravi/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Building an AWS Enterprise Landing Zone with Terraform: A Complete Guide]]></title>
            <link>https://lkravi.medium.com/building-an-aws-enterprise-landing-zone-with-terraform-a-complete-guide-4d79c7ec9eb1?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/4d79c7ec9eb1</guid>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[organization]]></category>
            <category><![CDATA[terraform]]></category>
            <category><![CDATA[aws-landing-zone]]></category>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Fri, 30 Jan 2026 10:18:55 GMT</pubDate>
            <atom:updated>2026-02-02T22:11:07.044Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A practical approach to setting up a secure, scalable multi-account AWS infrastructure</em></p><h4><strong>Introduction</strong></h4><p>When organizations grow, managing AWS resources in a single account becomes challenging. Security risks increase, cost tracking becomes difficult, and teams step on each other’s toes. The solution? A well-designed <strong>Landing Zone -</strong> a multi-account AWS environment with proper governance, security, and access management.</p><p>In this article, I’ll walk you through building an enterprise-grade AWS Landing Zone using Terraform. This is a real implementation that I built for a portfolio project, simulating a mid-size organization with around 500 employees.</p><p>Here is what we are going to discuss in this post:</p><p>- Setting up AWS Organizations with a multi-account strategy<br>- Designing Organizational Units (OUs) for different teams and environments<br>- Using Service Control Policies (SCPs) as security guardrails<br>- Configuring IAM Identity Center (SSO) for centralized access management<br>- Enabling security services like CloudTrail and GuardDuty across all accounts<br>- Structuring Terraform modules so the code stays clean and reusable</p><h4>Why Multi-Account Strategy?</h4><p>Before diving into the implementation, let me explain why we need multiple AWS accounts in the first place.</p><p>With a single AWS account, one misconfiguration can affect everything. Cost tracking becomes hard because all teams share the same billing. There are no real security boundaries between environments, and IAM policies become unmanageable as the team grows. On top of that, all teams share the same service quotas.</p><p>A multi-account setup solves all of these problems. Production is separated from development. If one account gets compromised, others remain safe. Each account has its own billing, making cost tracking straightforward. Compliance becomes easier because you can enforce rules per account. And teams get the autonomy to manage their own accounts without affecting others.</p><h4>Architecture Overview</h4><p>Here is the high-level architecture of our Landing Zone:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/718/1*jBOoHzDRnt9e35IkCk5khA.png" /></figure><p>Account Structure</p><pre>Root (Management Account)<br>  ├── Security OU<br>  │     ├── security-tooling (GuardDuty, Security Hub admin)<br>  │     └── log-archive (Centralized logs)<br>  ├── Infrastructure OU<br>  │     ├── network-hub (Transit Gateway, DNS)<br>  │     └── shared-services (CI/CD, Container Registry)<br>  ├── Workloads OU<br>  │     ├── Production OU<br>  │     │     ├── prod-engineering<br>  │     │     └── prod-data<br>  │     ├── Staging OU<br>  │     │     ├── dev-engineering<br>  │     │     └── staging-engineering<br>  │     └── Experimental OU<br>  │           └── sandbox-engineering<br>  └── Suspended OU (for decommissioned accounts)</pre><p>Design Decisions</p><p>Why this structure?</p><p>The Security OU sits at the root level because security accounts need organization-wide visibility. Infrastructure has its own OU because shared services should not mix with workloads. The Workloads OU has nested OUs so we can apply different policies for production versus non-production. And the Suspended OU gives us a safe place to move accounts before closing them.</p><p>Repository Structure</p><p>A well-organized repository is crucial for maintainability. Here is the structure I used:</p><pre>aws_devops/<br>  terraform/<br>    environments/<br>      management/<br>        main.tf<br>        terraform.tfvars.example<br>    modules/<br>      organization/<br>        main.tf<br>        scps.tf<br>        variables.tf<br>        outputs.tf<br>      identity-center/<br>        main.tf<br>        assignments.tf<br>        users.tf<br>        variables.tf<br>        outputs.tf<br>      security/<br>        cloudtrail/<br>        guardduty/</pre><p>The environments/ directory holds account-specific configurations. The modules/ directory contains reusable Terraform modules. Modules should be reusable and environment-agnostic. Environment directories contain only the specific configuration for each account.</p><p>Implementation Deep Dive</p><p><strong>Step 1: AWS Organizations Setup</strong></p><p>The foundation of our Landing Zone is AWS Organizations. It allows centralized management of multiple accounts.</p><pre>resource &quot;aws_organizations_organization&quot; &quot;this&quot; {<br>  aws_service_access_principals = [<br>    &quot;access-analyzer.amazonaws.com&quot;,<br>    &quot;cloudtrail.amazonaws.com&quot;,<br>    &quot;config.amazonaws.com&quot;,<br>    &quot;guardduty.amazonaws.com&quot;,<br>    &quot;sso.amazonaws.com&quot;,<br>  ]<br><br>  enabled_policy_types = [<br>    &quot;SERVICE_CONTROL_POLICY&quot;,<br>    &quot;TAG_POLICY&quot;,<br>    &quot;BACKUP_POLICY&quot;,<br>  ]<br><br>  feature_set = &quot;ALL&quot;<br>}</pre><p>A few important points here. The aws_service_access_principals list defines which AWS services can operate across your organization. The enabled_policy_types controls what types of policies you can attach to OUs. And feature_set = “ALL” enables all organization features, which is required for SCPs.</p><p><strong>Step 2: Organizational Units</strong></p><p>OUs are containers for accounts. They help you apply policies to groups of accounts.</p><pre>resource &quot;aws_organizations_organizational_unit&quot; &quot;security&quot; {<br>  name      = &quot;Security&quot;<br>  parent_id = aws_organizations_organization.this.roots[0].id<br><br>  tags = {<br>    Description = &quot;Security and compliance accounts&quot;<br>    ManagedBy   = &quot;terraform&quot;<br>  }<br>}<br><br>resource &quot;aws_organizations_organizational_unit&quot; &quot;workloads&quot; {<br>  name      = &quot;Workloads&quot;<br>  parent_id = aws_organizations_organization.this.roots[0].id<br>}<br><br>resource &quot;aws_organizations_organizational_unit&quot; &quot;production&quot; {<br>  name      = &quot;Production&quot;<br>  parent_id = aws_organizations_organizational_unit.workloads.id<br>}</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/577/1*tqzef1BM5UHbvv4bNDwGbQ.png" /></figure><p><strong>Step 3: Member Accounts</strong></p><p>Creating accounts is straightforward, but there is a catch — AWS accounts cannot be deleted via Terraform. They must be manually closed through the AWS Console.</p><pre>resource &quot;aws_organizations_account&quot; &quot;security_tooling&quot; {<br>  count = var.create_accounts ? 1 : 0<br><br>  name      = &quot;${var.organization_name}-security-tooling&quot;<br>  email     = var.account_emails[&quot;security_tooling&quot;]<br>  parent_id = aws_organizations_organizational_unit.security.id<br><br>  iam_user_access_to_billing = &quot;DENY&quot;<br>  role_name                  = &quot;OrganizationAccountAccessRole&quot;<br><br>  lifecycle {<br>    ignore_changes = [role_name, iam_user_access_to_billing]<br>  }<br>}</pre><p>Notice the count variable. This lets you set up the organization structure first, then create accounts in a separate step. This is important because account creation is a one-way operation.</p><h4>Email Address Strategy</h4><p>Each AWS account needs a unique email address. For testing, I used Gmail’s alias feature:</p><p>Base email: yourname@gmail.com</p><p>Generated aliases:<br> — yourname+aws-security@gmail.com<br> — yourname+aws-logs@gmail.com<br> — yourname+aws-network@gmail.com</p><p>All emails arrive in your single inbox. Here is how I configured this in Terraform:</p><pre>locals {<br>  email_parts  = split(&quot;@&quot;, var.base_email)<br>  email_user   = local.email_parts[0]<br>  email_domain = local.email_parts[1]<br><br>  generated_emails = {<br>    security_tooling = &quot;${local.email_user}+aws-security@${local.email_domain}&quot;<br>    log_archive      = &quot;${local.email_user}+aws-logs@${local.email_domain}&quot;<br>    network_hub      = &quot;${local.email_user}+aws-network@${local.email_domain}&quot;<br>  }<br>}</pre><p><strong>Step 4: Service Control Policies (SCPs)</strong></p><p>SCPs are guardrails that limit what actions can be performed in member accounts. They are attached to OUs and apply to all accounts within that OU.</p><pre>resource &quot;aws_organizations_policy&quot; &quot;deny_root_user&quot; {<br>  name        = &quot;DenyRootUser&quot;<br>  description = &quot;Deny all actions for root user in member accounts&quot;<br>  type        = &quot;SERVICE_CONTROL_POLICY&quot;<br><br>  content = jsonencode({<br>    Version = &quot;2012-10-17&quot;<br>    Statement = [<br>      {<br>        Sid       = &quot;DenyRootUser&quot;<br>        Effect    = &quot;Deny&quot;<br>        Action    = &quot;*&quot;<br>        Resource  = &quot;*&quot;<br>        Condition = {<br>          StringLike = {<br>            &quot;aws:PrincipalArn&quot; = &quot;arn:aws:iam::*:root&quot;<br>          }<br>        }<br>      }<br>    ]<br>  })<br>}</pre><p>Here are the SCPs I implemented:</p><p>- DenyRootUser — blocks root user actions, attached to Workloads OU<br>- DenyLeaveOrganization — prevents accounts from leaving, attached to Root<br>- RegionRestriction — allows only specific AWS regions, attached to Root<br>- DenyPublicS3 — blocks public S3 buckets, attached to Workloads OU<br>- ProtectSecurityServices — prevents disabling security tools, attached to Workloads and Security OUs<br>- SandboxRestrictions — limits expensive services, attached to Experimental OU</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/628/1*Bga_y_-juoeSBsocOuT5yA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/584/1*qDVAIX44XlDN26IYSYPraQ.png" /></figure><p><strong>Step 5: IAM Identity Center (SSO)</strong></p><p>IAM Identity Center provides centralized access management. Users log in once and can access multiple accounts with different permission levels.</p><p>There are three main components:</p><p>1. Permission Sets — define what users can do (for example, AdministratorAccess or ReadOnly)<br>2. Groups — organize users by role<br>3. Account Assignments — connect groups to accounts with specific permission sets</p><pre>resource &quot;aws_ssoadmin_permission_set&quot; &quot;administrator&quot; {<br>  name             = &quot;AdministratorAccess&quot;<br>  description      = &quot;Full administrator access&quot;<br>  instance_arn     = local.sso_instance_arn<br>  session_duration = &quot;PT4H&quot;<br>}<br><br>resource &quot;aws_ssoadmin_managed_policy_attachment&quot; &quot;administrator&quot; {<br>  instance_arn       = local.sso_instance_arn<br>  managed_policy_arn = &quot;arn:aws:iam::aws:policy/AdministratorAccess&quot;<br>  permission_set_arn = aws_ssoadmin_permission_set.administrator.arn<br>}</pre><p>Groups are created in the Identity Store and then assigned to accounts:</p><pre>resource &quot;aws_identitystore_group&quot; &quot;platform_admins&quot; {<br>  display_name      = &quot;Platform-Admins&quot;<br>  description       = &quot;Platform administrators with full access&quot;<br>  identity_store_id = local.identity_store_id<br>}<br><br>resource &quot;aws_ssoadmin_account_assignment&quot; &quot;platform_admins_management&quot; {<br>  instance_arn       = local.sso_instance_arn<br>  permission_set_arn = aws_ssoadmin_permission_set.administrator.arn<br><br>  principal_id   = aws_identitystore_group.platform_admins.group_id<br>  principal_type = &quot;GROUP&quot;<br><br>  target_id   = var.management_account_id<br>  target_type = &quot;AWS_ACCOUNT&quot;<br>}</pre><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*KkG135QwIarTqV4vJXgu_g.png" /></figure><p>Access Matrix</p><p>Here is how I mapped groups to accounts:</p><ul><li>Platform-Admins get Administrator access to all accounts</li><li>Security-Team gets Administrator access to security accounts and SecurityAudit access to all accounts</li><li>Engineering-Leads get PowerUser access to non-prod accounts and ReadOnly access to prod accounts</li><li>Engineering-Developers get Developer access to dev and sandbox accounts</li><li>DataScience-Team gets DataScientist access to the prod-data account</li><li>Auditors get ReadOnly access to all accounts</li></ul><p>Identity Source Options — What to Use in Production</p><p>In this project, I created demo users directly in IAM Identity Center for testing purposes. However, in a real production environment, you would not manage users this way. There are three identity source options available in AWS, and the right choice depends on your organization.</p><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*07c3uCjB24uKZworzGhNpg.png" /></figure><p><strong>The first</strong> option is the IAM Identity Center built-in directory. This is what I used for this demo. It is the simplest to set up and works well for small teams, testing, and proof-of-concept projects. You create and manage users directly in the IAM Identity Center console or via Terraform.</p><pre>resource &quot;aws_identitystore_user&quot; &quot;demo&quot; {<br>  identity_store_id = local.identity_store_id<br>  display_name      = &quot;Platform Admin&quot;<br>  user_name         = &quot;admin&quot;<br><br>  emails {<br>    value   = &quot;user+sso-admin@gmail.com&quot;<br>    primary = true<br>  }<br>}</pre><p><strong>The second</strong> option is Active Directory. If your organization already uses Microsoft Active Directory, you can connect it to IAM Identity Center. This works with both on-premises AD (through AD Connector) and AWS Managed Microsoft AD. Users authenticate with their existing AD credentials, groups sync automatically, and password policies are managed centrally in AD.</p><p>On-premises AD → AD Connector → IAM Identity Center<br>or<br>AWS Managed Microsoft AD → IAM Identity Center</p><p><strong>The third</strong> option is an External Identity Provider (IdP). This is for organizations already using identity providers like Okta, Azure AD (Entra ID), Google Workspace, or OneLogin. The integration works through SAML 2.0 for authentication and SCIM for automatic user provisioning. This means when you add or remove a user in Okta, the change automatically reflects in IAM Identity Center.</p><p>External IdP (Okta / Azure AD)<br> | — SAML 2.0 → IAM Identity Center (authentication)<br> | — SCIM → IAM Identity Center (user provisioning)</p><p>For most enterprises, I recommend the External Identity Provider approach because:</p><p>1. Single source of truth — manage users in one place<br>2. Automated provisioning — SCIM handles user creation and deletion<br>3. Consistent MFA — same multi-factor authentication across all applications<br>4. Compliance — easier audit trail for user access<br>5. Offboarding — disable a user in the IdP and access is revoked everywhere</p><p>The good news is that the Terraform configuration for groups, permission sets, and account assignments remains the same regardless of which identity source you choose. Only the user creation part changes.</p><p><strong>Step 6: Security Services</strong></p><p>CloudTrail</p><p>CloudTrail logs all API calls across your organization. This is essential for security auditing and compliance.</p><pre>resource &quot;aws_cloudtrail&quot; &quot;organization&quot; {<br>  name                       = &quot;${var.name_prefix}-organization-trail&quot;<br>  s3_bucket_name             = aws_s3_bucket.cloudtrail.id<br>  is_organization_trail      = true<br>  is_multi_region_trail      = true<br>  enable_log_file_validation = true<br>  kms_key_id                 = aws_kms_key.cloudtrail.arn<br><br>  event_selector {<br>    read_write_type           = &quot;All&quot;<br>    include_management_events = true<br>  }<br>}</pre><p>Setting is_organization_trail = true means this single trail captures API calls from all accounts in the organization. All logs are stored in a centralized S3 bucket with KMS encryption.</p><p>GuardDuty</p><p>GuardDuty is a threat detection service. It analyzes CloudTrail logs, VPC Flow Logs, and DNS logs to identify threats.</p><pre>resource &quot;aws_guardduty_detector&quot; &quot;primary&quot; {<br>  enable                       = true<br>  finding_publishing_frequency = &quot;FIFTEEN_MINUTES&quot;<br><br>  datasources {<br>    s3_logs {<br>      enable = true<br>    }<br>    kubernetes {<br>      audit_logs {<br>        enable = true<br>      }<br>    }<br>    malware_protection {<br>      scan_ec2_instance_with_findings {<br>        ebs_volumes {<br>          enable = true<br>        }<br>      }<br>    }<br>  }<br>}<br><br>resource &quot;aws_guardduty_organization_admin_account&quot; &quot;delegated_admin&quot; {<br>  count            = var.enable_delegated_admin ? 1 : 0<br>  admin_account_id = var.delegated_admin_account_id<br>}</pre><p>Delegating GuardDuty administration to the security account means the security team can manage findings without needing access to the management account.</p><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*QMPgMAFdDym_dSrMAohATw.png" /></figure><h4>Challenges and Lessons Learned</h4><p>AWS Accounts Cannot Be Deleted</p><p>Once you create an AWS account through Organizations, you cannot delete it via Terraform. This caught me off guard. If you need to recover, your options are: remove the account from Terraform state using terraform state rm, move the account to the Suspended OU, and eventually close the account manually through the AWS Console.</p><p>IAM Identity Center Must Be Enabled Manually</p><p>IAM Identity Center cannot be enabled through Terraform. You have to go to the AWS Console and click Enable first. I handled this by wrapping the module in a conditional:</p><pre>module &quot;identity_center&quot; {<br>  source = &quot;../../modules/identity-center&quot;<br>  count  = var.enable_identity_center ? 1 : 0<br>}</pre><p>This way, you deploy the Landing Zone in phases. First without Identity Center, enable it in the console, then set the variable to true and apply again.</p><h4>Deployment Steps</h4><p>Here is the order I followed to deploy the Landing Zone. The phased approach is important because some resources depend on others, and some require manual steps.</p><p><strong>Phase 1: Foundation</strong></p><p>Create a management account manually in AWS. Run terraform init. Apply with create_accounts = false. This creates the organization, OUs, and SCPs first.</p><p><strong>Phase 2: Enable SSO</strong></p><p>Go to AWS Console, navigate to IAM Identity Center, and click Enable. This step cannot be automated.</p><p><strong>Phase 3: Create Accounts and Identity Center</strong></p><p>Set create_accounts = true and enable_identity_center = true. Run terraform apply. This creates all 9 member accounts, sets up permission sets, groups, and account assignments.</p><p><strong>Phase 4: Create Demo Users</strong></p><p>Set create_demo_users = true. Run terraform apply. Then go to IAM Identity Center in the console and send password reset emails to each user.</p><h4>Key Takeaways</h4><ul><li>Start with design. Plan your OU structure before writing any code.</li><li>Use modules. Keep Terraform code reusable and maintainable.</li><li>Implement SCPs early. Guardrails prevent issues before they happen.</li><li>Centralize security. Dedicated security accounts simplify auditing.</li><li>Plan for the unexpected. AWS accounts cannot be deleted easily, so think before you create.</li></ul><p><a href="https://github.com/lkravi/aws_landing_zone">GitHub - lkravi/aws_landing_zone</a></p><p>Resources</p><ul><li>AWS Organizations Documentation<br><a href="https://docs.aws.amazon.com/organizations/">docs.aws.amazon.com/organizations</a></li><li>AWS Landing Zone Accelerator<br><a href="https://aws.amazon.com/solutions/implementations/landing-zone-accelerator-on-aws/">aws.amazon.com/solutions/implementations/landing-zone-accelerator-on-aws</a></li><li>Terraform AWS Provider<br><a href="https://registry.terraform.io/providers/hashicorp/aws/latest">registry.terraform.io/providers/hashicorp/aws/latest</a></li><li>AWS Well-Architected Framework<br><a href="https://aws.amazon.com/architecture/well-architected/">aws.amazon.com/architecture/well-architected</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4d79c7ec9eb1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AWS DevOps Agent: Your Autonomous On-Call Engineer is Here (Preview)]]></title>
            <link>https://lkravi.medium.com/aws-devops-agent-your-autonomous-on-call-engineer-is-here-preview-8d220045e620?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/8d220045e620</guid>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[aiops]]></category>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[incident-response]]></category>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Sun, 11 Jan 2026 12:05:07 GMT</pubDate>
            <atom:updated>2026-01-11T12:05:07.385Z</atom:updated>
            <content:encoded><![CDATA[<p>As engineers, we all know the pain of production incidents. You get an alert, then spend hours checking CloudWatch dashboards, reading logs, reviewing deployment history, and updating Slack threads. It takes a lot of time and effort.</p><p>AWS has announced something interesting to help with this problem — <strong>AWS DevOps Agent</strong>, now available in public preview.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cFyefC0080pf_uhSq-N3XQ.png" /></figure><h3>What is AWS DevOps Agent?</h3><p>AWS DevOps Agent is what AWS calls a <strong>“frontier agent”</strong> — a new type of AI agent that works autonomously. It can run for hours or even days without needing constant human input.</p><p>Think of it as a virtual on-call engineer that works 24/7.</p><p>When issues happen, it automatically:</p><ul><li><strong>Correlates data across your operational tools</strong> — metrics, logs, and code deployments from GitHub or GitLab</li><li><strong>Identifies root causes</strong> and recommends solutions</li><li><strong>Manages incident communication</strong> using Slack channels</li><li><strong>Maintains detailed investigation timelines</strong></li><li><strong>Analyzes past incidents</strong> to help prevent future issues</li></ul><p>The agent builds an <strong>intelligent application topology</strong> — a map of your system components, their relationships, and deployment history. This helps it find deployment-related problems during investigations.</p><h3>Integrations</h3><p>One of the best things about AWS DevOps Agent is that it works with many popular tools, not just AWS services.</p><p><strong>Observability Platforms:</strong></p><ul><li>Amazon CloudWatch</li><li>Datadog</li><li>Dynatrace</li><li>New Relic</li><li>Splunk</li></ul><p><strong>CI/CD &amp; Source Control:</strong></p><ul><li>GitHub Actions</li><li>GitLab CI/CD</li></ul><p><strong>Incident Management:</strong></p><ul><li>ServiceNow (built-in support)</li><li>PagerDuty (via webhooks)</li><li>Other tools via configurable webhooks</li></ul><p><strong>Bring Your Own Tools:</strong> Using the <strong>BYO Model Context Protocol (MCP) server</strong> feature, you can connect additional tools like Grafana, Prometheus, or your own custom tools.</p><p>This flexibility means you can use AWS DevOps Agent with your existing tools without changing your setup.</p><h3>Getting Started with AWS DevOps Agent</h3><p>Let me walk you through how to set this up.</p><h3>Step 1: Create an Agent Space</h3><p>An <strong>Agent Space</strong> defines the scope of what AWS DevOps Agent can access as it performs tasks. Think of it as a boundary for your investigations.</p><p>You can organize Agent Spaces based on your operational model:</p><ul><li>One Agent Space per application</li><li>One per on-call team managing multiple services</li><li>Centralized approach for the entire organization</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cu3Km3Dw77mhCNdkXOMBMg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i0_CAi4mXpoCF_caxxw13w.png" /></figure><p>Navigate to the <strong>AWS DevOps Agent</strong> section in the AWS Management Console and select <strong>Create Agent Space</strong>.<br>Enter a name for your space and create the IAM roles it uses to introspect AWS resources. You can also enable the <strong>AWS DevOps Agent web app</strong> at this stage (or later).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wlFoZFpchRGr2gktVDdyPQ.png" /></figure><p>After creation, check the <strong>Topology</strong> tab. This shows the key resources, entities, and relationships the agent has identified as relevant for performing its tasks. As your agent completes more investigations, it discovers and adds new resources to this topology.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RV6bc9WFwLchxoW_mTB3oQ.png" /></figure><h3>Step 2: Configure the Web App for Operators</h3><p>The AWS DevOps Agent web app provides a web interface for on-call engineers to:</p><ul><li>Manually trigger investigations</li><li>View investigation details, including relevant topology elements</li><li>Steer investigations with additional context</li><li>Ask questions about ongoing investigations</li></ul><p>You can access the web app directly from your Agent Space in the console via the <strong>Operator access</strong> link. For team access, you can use <strong>AWS IAM Identity Center</strong> to manage users and groups.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4yx_33xF5lCyQOElKqnh0w.png" /></figure><h3>Step 3: Start an Investigation</h3><p>When an incident occurs, you can start an investigation in several ways:</p><ul><li><strong>Automatically</strong> — Configure the agent to respond to alerts from ServiceNow, PagerDuty, or other ticketing systems</li><li><strong>Manually</strong> — Click <strong>Start Investigation</strong> in the web app</li></ul><p>The web app offers preconfigured starting points:</p><ul><li><strong>Latest alarm</strong> — Investigate your most recent triggered alarm</li><li><strong>High CPU usage</strong> — Investigate high CPU utilization across compute resources</li><li><strong>Error rate spike</strong> — Investigate recent increases in application error rates</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YUQKrNPt59HvTDQlyr_1ww.png" /></figure><p>Enter the investigation details, starting point, date/time of the incident, and the AWS Account ID.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/673/1*Kg5AstIYKXnm4zhX2Hcl4w.png" /></figure><h3>Step 4: Watch the Investigation</h3><p>Here you can see the agent working. In the web app, you can watch the investigation happen <strong>in real time</strong>.</p><p>The agent:</p><ol><li>Identifies the application stack</li><li>Correlates metrics from CloudWatch (or external tools like Splunk)</li><li>Checks logs from CloudWatch Logs</li><li>Reviews recent code changes from GitHub/GitLab</li><li>Analyzes traces from AWS X-Ray</li><li>Identifies error patterns</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TUCeosrRpWXQAaEwSKObqw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ExKioSixq78Y6ZFcZpBAdQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sXoI3HuhZvuaDnXVlXtt9Q.png" /></figure><h3>Step 5: Review Findings and Root Cause</h3><p>The agent provides a detailed investigation summary with:</p><ul><li>Timeline of events leading to the alarm</li><li>Identified error patterns</li><li>Root cause analysis</li><li>Suggested monitoring improvements</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oXe5mvYBU2rYHHeKYvW__g.png" /></figure><p>Through the <strong>investigation chat interface</strong>, you can:</p><ul><li>Ask clarifying questions like <em>“which logs did you analyze?”</em></li><li>Steer the investigation: <em>“focus on these specific log groups and rerun your analysis”</em></li><li>Create an <strong>AWS Support case with a single click</strong> — automatically populated with the agent’s findings</li></ul><h3>Beyond Incident Response: Proactive Improvements</h3><p>AWS DevOps Agent does not only help during incidents — it also helps you prevent future problems.</p><p><strong>During Active Incidents:</strong> The agent provides <strong>mitigation plans</strong> through its incident mitigations tab. These include step-by-step guidance for developers. It can also work with development tools like <strong>Kiro</strong>.</p><p><strong>For Long-Term Improvements:</strong> The agent analyzes your past incidents and suggests improvements for:</p><ul><li>Better observability setup</li><li>Infrastructure configuration fixes</li><li>Deployment pipeline improvements</li><li>Multi-AZ deployment recommendations</li></ul><p>According to AWS, in a future release, the agent will also analyze code bugs and testing coverage.</p><h3>Key Features Summary</h3><p>Feature Description <strong>Frontier Agent Architecture</strong> Autonomous, works for hours/days without intervention <strong>Multi-Tool Integration</strong> CloudWatch, Datadog, Dynatrace, New Relic, Splunk <strong>CI/CD Correlation</strong> GitHub Actions, GitLab CI/CD deployment tracking <strong>BYO MCP Server</strong> Integrate custom tools, Grafana, Prometheus <strong>Application Topology</strong> Automatic mapping of components and relationships <strong>Slack Integration</strong> Real-time stakeholder updates <strong>Investigation Chat</strong> Interactive Q&amp;A with the agent <strong>One-Click Support Case</strong> AWS Support integration with auto-populated findings <strong>Proactive Recommendations</strong> Prevent future incidents through systematic analysis</p><h3>Availability and Pricing</h3><p><strong>Region:</strong> US East (N. Virginia) — but can monitor applications deployed in <strong>any Region</strong> across <strong>multiple AWS accounts</strong>.</p><p><strong>Pricing:</strong> During the preview period, AWS DevOps Agent is available <strong>at no charge</strong>, with a limit on agent task hours per month.</p><p>To get started, visit the <a href="https://aws.amazon.com/devops-agent/">AWS DevOps Agent page</a>.</p><h3>My Initial Thoughts</h3><p>Here is what I think about AWS DevOps Agent:</p><p><strong>What I Like:</strong></p><ul><li>The <strong>BYO MCP server</strong> feature lets you connect tools outside AWS</li><li><strong>Automatic deployment correlation</strong> from GitHub/GitLab saves investigation time</li><li>The <strong>intelligent topology</strong> learns your infrastructure over time</li><li><strong>Slack integration</strong> keeps stakeholders updated during incidents</li><li><strong>One-click AWS Support case</strong> with investigation details already filled in</li></ul><p><strong>Things to Consider:</strong></p><ul><li>Currently only runs in us-east-1 (but can monitor resources in any region)</li><li>This is a preview, so features may change</li><li>Results depend on how good your observability data is</li><li>The recommendations feature needs more incidents to give useful suggestions</li></ul><h3>Conclusion</h3><p>AWS DevOps Agent is an interesting step toward AI-assisted operations.</p><p>It combines operational insights, multi-tool integration, and practical recommendations. This can help teams move from <strong>reactive problem-solving to proactive system improvement</strong>.</p><p>If you run workloads on AWS and spend a lot of time investigating incidents, I recommend trying this during the preview period.</p><p>Have you tried AWS DevOps Agent? Share your experience in the comments!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8d220045e620" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When the Cloud Sneezes: Lessons from the Recent AWS Outage]]></title>
            <link>https://lkravi.medium.com/when-the-cloud-sneezes-lessons-from-the-recent-aws-outage-e6b16536f9df?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/e6b16536f9df</guid>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[multi-cloud]]></category>
            <category><![CDATA[cloud-architecture]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[high-availability]]></category>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Thu, 23 Oct 2025 21:27:24 GMT</pubDate>
            <atom:updated>2025-10-23T21:27:24.087Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*89hQpD9A_fyKHluy-AFrTA.png" /></figure><p>A couple of days ago, AWS had a one of those “bad hair days” again, this time in us-east-1, the busiest and famous region in their ecosystem.</p><p>What started as a small internal issue in <strong>DynamoDB’s DNS automation system</strong> quickly turned into a domino effect that affected <strong>EC2</strong>, <strong>Lambda</strong>, <strong>Load Balancers</strong>, and even <strong>IAM sign-ins</strong>.</p><p>In simple terms: a small automation script pushed an old DNS plan, removed some critical records, and boom — services couldn’t find each other. DNS went blind for a while, and when DNS breaks, <em>everything breaks</em>.</p><p>The outage lasted hours. Many customers saw failed deployments, stuck EC2 launches, and unavailable APIs.</p><p>And yes — it reminded us again that <strong>“the cloud” is just someone else’s computer</strong>… and that someone can have a bad day too.</p><h3>What can we learn from this as AWS users?</h3><p>AWS is incredibly resilient overall — but this incident reminds us of something important: <strong>even the most reliable cloud platform can fail regionally</strong>.</p><p>So, as customers, we need to ask ourselves a few honest questions:</p><ul><li>Do our applications survive when one region goes dark?</li><li>Is the cost of regional high availability justified for what we do?</li><li>Could multi-cloud really help us? Or will it just multiply our headaches?</li></ul><p>Let’s unpack those.</p><h3>Regional High Availability — Great in Theory, Expensive in Practice</h3><p>Running across multiple AWS regions (say, Sydney + Singapore) sounds great — until you open the bill.</p><p>You’ll pay for:</p><ul><li>Double infrastructure (EC2, RDS, networking, etc.)</li><li>Cross-region data transfer (which isn’t cheap)</li><li>Added complexity in replication, routing, and failover testing</li></ul><p>Most startups and even mid-size businesses find regional HA too costly or complex for their current risk appetite.</p><p>So what happens?<br> They stay in one region, maybe with multi-AZ setup — and hope AWS keeps the lights on.</p><p>Honestly, that’s not a bad decision <em>if</em> you’ve done the math.</p><p>If your service can handle a few hours of downtime within your <strong>error budget</strong>, staying single-region might actually be the smart, lean choice.</p><p>Reliability is always a trade-off between <strong>cost, complexity, and impact</strong>.</p><h3>Multi-Cloud — The Shiny but Heavy Option</h3><p>Ah, multi-cloud — the dream of every CIO who’s been burned once by a single vendor.</p><p>In theory, running across AWS + GCP + Azure sounds like ultimate resilience. But reality hits hard:</p><ul><li><strong>Cost:</strong> You pay double or triple infra.</li><li><strong>Skill:</strong> You need engineers fluent in all three clouds. (That’s a rare species.)</li><li><strong>Feature gaps:</strong> Each cloud does things differently — networking, IAM, storage — nothing is truly portable.</li><li><strong>Complexity:</strong> CI/CD, secrets, observability, and failover logic become a nightmare.</li></ul><p>So yes, multi-cloud <em>can</em> protect you from regional or provider-wide issues — but only if your organization is mature enough to handle that operational overhead.</p><p>For most businesses, it’s like buying three luxury cars when you only need one — great bragging rights, but not the best ROI.</p><h3>So… What’s the Realistic Middle Ground?</h3><p>Now, this “middle ground” idea isn’t one-size-fits-all.</p><p>Some systems — like <strong>mission-critical banking apps, trading platforms, healthcare systems</strong>, or <strong>national-scale government services</strong> — can’t afford regional outages. By design, they <strong>must</strong> go for <strong>regional HA</strong>, <strong>multi-cloud</strong>, or even <strong>hybrid</strong> models that combine on-premise control with cloud scalability.</p><p>And that’s perfectly right.</p><p>But for <em>most</em> of us — startups, SaaS platforms, internal tools, data pipelines, or event apps — we don’t need that extreme level of redundancy. What we really need is <strong>resilience without over-engineering</strong>.</p><p>Here’s what that middle ground often looks like:</p><p>✅ <strong>Multi-AZ setup</strong> — gives solid fault tolerance within a region without doubling cost.<br> ✅ <strong>Graceful degradation</strong> — keep core functions running even if a dependency fails.<br> ✅ <strong>Caching and retries</strong> — so DNS or API hiccups don’t crash the whole stack.<br> ✅ <strong>Clear error budget</strong> — know how much downtime you can live with before you spend big on HA.<br> ✅ <strong>Operational readiness</strong> — monitoring, alerting, and simulated failure (game days).</p><p>The truth is, not every workload needs to survive a regional apocalypse. Some just need to bounce back quickly and keep customer trust intact.</p><p>Building the <em>right amount</em> of reliability is a skill — not a checkbox.</p><h3>My Takeaway</h3><p>The AWS outage wasn’t about a bad engineer or a broken script — it was about how small automation bugs in complex systems can ripple into massive failures.</p><p>As AWS users, we can’t control their internals, but we <em>can</em> control how our workloads respond to chaos.</p><p>Architect smartly, stay humble, and accept that 100% uptime is a myth — but <strong>well-architected systems fail gracefully</strong>.</p><h3>References</h3><ul><li><a href="https://aws.amazon.com/message/101925/?utm_source=chatgpt.com">AWS Service Event: October 19–20, 2025 — us-east-1 Region</a></li></ul><p>🤖 AI Tools Used<br><em>This blog post was formatted with the help of </em><strong><em>ChatGPT (GPT-5)</em></strong><em>, and the featured image was generated using </em><strong><em>OpenAI image generation tools</em></strong><em> based on the concept “When the Cloud Sneezes”.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e6b16536f9df" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Simple Istio Annotation Trick That Can Save You Hours of Debugging ]]></title>
            <link>https://lkravi.medium.com/a-simple-istio-annotation-trick-that-can-save-you-hours-of-debugging-33265d799b89?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/33265d799b89</guid>
            <category><![CDATA[istio]]></category>
            <category><![CDATA[annotations]]></category>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Wed, 20 Aug 2025 02:00:35 GMT</pubDate>
            <atom:updated>2025-08-20T02:10:49.482Z</atom:updated>
            <content:encoded><![CDATA[<h3>A Simple Istio Annotation Trick That Can Save You Hours of Debugging</h3><p>If you work with Istio regularly, you’ve probably run into those <em>maddeningly misleading</em> connectivity errors — everything looks fine in Kubernetes, but the app can’t reach an external service.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*eCgn6CvhInlicmfx" /></figure><p>Is it the <strong>DestinationRule?</strong><br> Is it a missing <strong>ServiceEntry?</strong><br> Is it your k8 <strong>NetworkPolicy?</strong></p><p>Before jumping into full-blown Istio resource forensics, there’s a quick way to eliminate Istio itself from the equation — <strong>by temporarily excluding specific outbound traffic from the sidecar proxy.</strong></p><p>👉 Two annotations that can save your day:</p><pre>traffic.sidecar.istio.io/excludeOutboundIPRanges: &quot;&lt;IP/CIDR&gt;&quot;<br>traffic.sidecar.istio.io/excludeOutboundPorts: &quot;&lt;port-list&gt;&quot;</pre><h4>When to use this</h4><p>Use these annotations on your Deployment/Pod when you suspect Istio might be intercepting traffic (envoy sidecar) and want to confirm if the underlying <strong>network path (K8s → Route Table → Firewall → Target)</strong> is actually working.</p><p>For example:</p><pre>annotations:<br>  traffic.sidecar.istio.io/excludeOutboundIPRanges: &quot;10.128.0.15/32&quot;<br>  traffic.sidecar.istio.io/excludeOutboundPorts: &quot;3306&quot;</pre><p>This tells Envoy to bypass the sidecar for traffic going to that IP or port — so your app can make a <strong>normal Kubernetes egress call</strong>, without Istio interfering.</p><h4>Debugging Workflow</h4><ol><li>Add the annotation to <strong>exclude</strong> your target IP/port.</li><li>Test the connectivity from the pod.</li></ol><p>* If it works ⬅️ → the problem is with your <strong>Istio config</strong> (DestinationRule, ServiceEntry, TLS mode, etc.)</p><p>* If it still fails ⬅️ → the issue is somewhere in the <strong>networking layer</strong> outside of Istio.</p><p>3. Once confirmed, remove the annotation and fix the actual Istio resource.</p><p>💡 <em>This small trick has personally saved me countless hours — turning Istio from a suspect into a verified ally during troubleshooting.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=33265d799b89" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Avoid Startup Failures in Istio: Wait for the Sidecar Proxy to Be Ready!]]></title>
            <link>https://lkravi.medium.com/avoid-startup-failures-in-istio-wait-for-the-sidecar-proxy-to-be-ready-05638624af4d?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/05638624af4d</guid>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Sun, 15 Jun 2025 14:15:48 GMT</pubDate>
            <atom:updated>2025-06-15T14:15:48.674Z</atom:updated>
            <content:encoded><![CDATA[<p>If you’ve ever deployed a Kubernetes app in an Istio-enabled environment and found your container failing during startup for seemingly no reason — <strong>you’re not alone</strong>.</p><p>One of the sneakiest issues in service meshes like Istio is when your app container starts before the istio-proxy sidecar is fully initialized. This leads to network calls like pulling from GitHub, calling APIs, or even DNS resolution silently <strong>failing during boot</strong>, only to magically work later when you kubectl exec into the pod.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*wbJZvIo5Hdhf8MyS" /></figure><h3>Problem</h3><p>Many applications perform <strong>network operations early in their lifecycle</strong> — cloning repos, calling APIs, downloading configs, etc. But when Istio is in the picture, <strong>Envoy needs to be ready first</strong>, since it intercepts all traffic.</p><p>If Envoy isn’t ready, your app is essentially stuck in a network blackhole, causing timeouts, restarts, or cryptic startup errors.</p><h3>The Fix Introduced in Istio 1.7: holdApplicationUntilProxyStarts</h3><p>To solve this, Istio 1.7 introduced a new feature:</p><pre>holdApplicationUntilProxyStarts: true</pre><p>This instructs the sidecar injector to:</p><ul><li>Inject the istio-proxy at the top of the pod&#39;s container list</li><li><strong>Delay all other containers</strong> from starting until the sidecar proxy is fully initialized and ready</li></ul><p>This avoids the need for hacky sleep commands and ensures your app has a functional network stack from the moment it starts.</p><h3>How to Use It</h3><h3>Option 1: Globally (via IstioOperator)</h3><p>In your IstioOperator manifest:</p><pre>spec:<br>  values:<br>    global:<br>      proxy:<br>        holdApplicationUntilProxyStarts: true</pre><p>Then install/update your control plane with:</p><pre>istioctl install -f istio-operator.yaml</pre><h3>Option 2: Per Pod (using Annotations)</h3><p>To enable it for a specific app only:</p><pre>metadata:<br>  annotations:<br>    proxy.istio.io/config: &#39;{ &quot;holdApplicationUntilProxyStarts&quot;: true }&#39;</pre><p>Example Deployment:</p><pre>apiVersion: apps/v1<br>kind: Deployment<br>metadata:<br>  name: git-cloner<br>  annotations:<br>    proxy.istio.io/config: &#39;{ &quot;holdApplicationUntilProxyStarts&quot;: true }&#39;</pre><h3>Result</h3><p>With this enabled, Istio:</p><ul><li>Starts the istio-proxy sidecar first</li><li>Waits for it to become healthy (/healthz/ready)</li><li><strong>Only then</strong> starts your application container</li></ul><p>This avoids mysterious connectivity issues and makes your pod startup more predictable and stable.</p><h3>Final Thoughts</h3><p>This setting has been available since <strong>Istio 1.7</strong>, yet it’s still underused in many clusters. If your workloads depend on network connectivity at startup, enabling holdApplicationUntilProxyStarts is a <strong>best practice</strong>.</p><p>It’s a one-line fix that can save hours of debugging.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=05638624af4d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Redis Cluster + Istio: A Networking Gotcha with GCP Memorystore Cluster mode]]></title>
            <link>https://lkravi.medium.com/redis-cluster-istio-a-networking-gotcha-with-gcp-memorystore-cluster-mode-53582b0cad67?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/53582b0cad67</guid>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Sun, 15 Jun 2025 12:05:24 GMT</pubDate>
            <atom:updated>2025-06-15T14:28:37.248Z</atom:updated>
            <content:encoded><![CDATA[<p>We recently encountered a subtle yet impactful networking issue while integrating Google Cloud Memorystore for Redis Cluster with an Istio-enabled GKE cluster. This post outlines the problem and the workaround we employed to resolve it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*l4L9yW26bwWPBlbt" /></figure><h3>The Issue</h3><p>Redis Cluster distributes data across multiple nodes, each responsible for a subset of the keyspace. Clients initially connect to a startup node (typically on port 6379) to retrieve the cluster topology. Subsequently, they communicate directly with the appropriate nodes responsible for specific key slots. These nodes can reside on different IPs and ports.</p><p>In our setup, the Redis cluster attempted to connect to a node on port 11006, resulting in the following error:</p><pre>Error connecting to 10.XXX.XXX.6:11006 [SSL: Unexpected EOF while reading]</pre><p>This suggests that the Redis cluster is trying to communicate with a cluster node on port 11006, which is being blocked by Istio due to the ServiceEntry not allowing traffic on that port.</p><h3>Why This Happens</h3><p>Redis Cluster relies on a <strong>dynamic port range</strong> for inter-node communication, typically from 11000 to 13047. Since this is a <strong>fully managed service by Google</strong>, we <strong>don’t have control over which ports are used</strong> or how they&#39;re assigned. The ports are picked dynamically and can vary with each connection or instance.</p><p>Unfortunately, Istio <strong>does not support port ranges in ServiceEntries</strong> (see: <a href="https://github.com/istio/istio/issues/23983">istio/istio#23983</a>). That means you would need to explicitly define each port individually — and you’re limited to 250 ports per ServiceEntry, which doesn’t scale well for Redis’ dynamic needs.</p><h3>Our Workaround</h3><p>To circumvent this limitation, we utilized Istio’s sidecar proxy annotations to exclude specific outbound ports and IPs from being intercepted by the Envoy proxy. Here’s how we did it:</p><ol><li><strong>Exclude Outbound Ports</strong>: We used the traffic.sidecar.istio.io/excludeOutboundPorts annotation to specify the range of ports used by Redis Cluster.</li><li><strong>Exclude Outbound IPs</strong>: We used the traffic.sidecar.istio.io/excludeOutboundIPRanges annotation to specify the IP ranges of the Redis Cluster nodes.</li></ol><p>By applying these annotations to our pods, we allowed traffic to Redis Cluster nodes to bypass the Istio proxy, enabling direct communication and resolving the connectivity issue.</p><h3>Key Takeaways</h3><ul><li><strong>Understand Istio’s Limitations</strong>: Be aware of the limitations of Istio’s ServiceEntry, especially regarding port range specifications.</li><li><strong>Use Sidecar Annotations</strong>: Utilize sidecar proxy annotations to exclude specific traffic from being intercepted by the proxy when necessary.</li><li><strong>Monitor and Validate</strong>: Always monitor your services and validate configurations to ensure that traffic flows as expected.</li></ul><ul><li><a href="https://github.com/istio/istio/issues/23983">Having port range or port wildcard in configurations and rules · Issue #23983 · istio/istio</a></li><li><a href="https://github.com/istio/istio/issues/55661">ServiceEntries addresses and ports limit to 256 items · Issue #55661 · istio/istio</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=53582b0cad67" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[EKS Auto mode]]></title>
            <link>https://awstip.com/eks-auto-mode-0149c684f77e?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/0149c684f77e</guid>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[terraform]]></category>
            <category><![CDATA[aws-eks]]></category>
            <category><![CDATA[aws]]></category>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Tue, 07 Jan 2025 14:43:34 GMT</pubDate>
            <atom:updated>2025-01-08T18:11:58.591Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/360/0*-CaZeid4mdwkpQc6.png" /></figure><p>AWS recently introduced <strong>EKS Auto Mode</strong>, a fully automated solution for managing compute, storage, and networking for Kubernetes clusters. This long-awaited feature significantly reduces the operational burden by offloading most of the cluster management tasks to AWS. It’s designed for individuals and organizations that face challenges in managing Kubernetes clusters due to limited expertise, bandwidth, or resources.</p><p>With <strong>EKS Auto Mode</strong>, the need for deep Kubernetes expertise, ongoing infrastructure management, and manual capacity planning is eliminated. AWS automatically selects the optimal EC2 instances for your workloads, ensuring cost optimization while delivering the right size and performance. This feature not only optimizes resource utilization but also seamlessly handles operating system patches and Kubernetes version upgrades, all without impacting your applications.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UZ650zz9R7WqLqthAmJ1VQ.png" /></figure><p>Additionally, <strong>EKS Auto Mode</strong> simplifies cluster operations by managing system components such as the kubelet, container runtime, and system agents—tasks that were previously the customer’s responsibility. This frees up users to focus on application development rather than infrastructure concerns.</p><p>For businesses looking to scale efficiently while maintaining robust application performance, this is a game-changer. EKS Auto Mode empowers teams to focus on innovation rather than infrastructure, making Kubernetes accessible even to those without deep technical expertise.</p><p>Let’s dive deeper into this feature and explore how it works to simplify Kubernetes cluster management while enhancing operational efficiency.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9nmVg1gwpfY_JlwgANRxsA.png" /></figure><p>To help demystify AWS EKS Auto Mode pricing, I’ve created this infographic based on the official example provided in the AWS EKS pricing documentation (aws.amazon.com/eks/pricing/). While AWS provides detailed pricing information, visualizing the cost breakdown can make it easier to understand how different components contribute to your total bill. This visualization walks through their example of running a three-tier application — consisting of a web frontend, backend services, and batch processing jobs — on EKS Auto Mode in the US West (Oregon) region. By breaking down the workload requirements, instance selection, and associated costs, we can better understand the pricing structure of EKS Auto Mode, including the base EC2 costs, management fees, and control plane charges. Let’s explore how the total monthly cost of $1,245.44 is calculated!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*80s2Vw-EhR5q0gY6YXWklA.png" /></figure><p>In this hands-on demonstration, we’ll explore AWS EKS Auto Mode in action. We’ll walk through three key aspects of managing a modern Kubernetes cluster:</p><p>1. Cluster Provisioning: Using Infrastructure as Code (IaC) with Terraform, we’ll automate the setup of an EKS cluster with Auto Mode enabled. This approach ensures our infrastructure is reproducible and follows best practices for cloud resource management.</p><p>2. Application Deployment and Scaling: We’ll deploy a sample application to see EKS Auto Mode in action. The interesting part here is that we won’t need to worry about manually managing worker nodes. Instead, we’ll observe how EKS Auto Mode automatically:<br> — Determines the right size and type of EC2 instances needed for our workload<br> — Scales the node count up when our application needs more resources<br> — Scales down when there’s less demand, helping optimize costs<br> — Distributes pods across multiple Availability Zones for high availability</p><p>3. Cluster Upgrade Process: Finally, we’ll perform a live version upgrade from Kubernetes 1.30 to 1.31. This is particularly interesting because we’ll see how EKS Auto Mode handles the upgrade process while:<br> — Maintaining application availability<br> — Managing node rotations<br> — Ensuring workload compatibility with the new version</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Fuj7i16EstFw%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Duj7i16EstFw&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fuj7i16EstFw%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/c2de2f4c2570f6e67df90fdd0ac37024/href">https://medium.com/media/c2de2f4c2570f6e67df90fdd0ac37024/href</a></iframe><p>This demonstration will give us practical insights into how EKS Auto Mode simplifies cluster management and handles common operational tasks that traditionally required significant manual intervention. Whether you’re a DevOps engineer, platform engineer, or just curious about modern container orchestration, this demo will show you the real-world benefits of using EKS Auto Mode.</p><p><a href="https://github.com/lkravi/eks_auto">GitHub - lkravi/eks_auto</a></p><p>References</p><p><a href="https://aws.amazon.com/eks/pricing/">https://aws.amazon.com/eks/pricing/</a><br><a href="https://docs.aws.amazon.com/eks/latest/userguide/automode.html">https://docs.aws.amazon.com/eks/latest/userguide/automode.html</a><br><a href="https://aws.amazon.com/eks/auto-mode/">https://aws.amazon.com/eks/auto-mode/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0149c684f77e" width="1" height="1" alt=""><hr><p><a href="https://awstip.com/eks-auto-mode-0149c684f77e">EKS Auto mode</a> was originally published in <a href="https://awstip.com/">AWS Tip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Integrated Testing to Scalable Cloud Deployment: A Complete Infrastructure Template]]></title>
            <link>https://lkravi.medium.com/integrated-testing-to-scalable-cloud-deployment-a-complete-infrastructure-template-df262c78abcf?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/df262c78abcf</guid>
            <category><![CDATA[fargate]]></category>
            <category><![CDATA[docker-compose]]></category>
            <category><![CDATA[containers]]></category>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Fri, 21 Jun 2024 08:23:43 GMT</pubDate>
            <atom:updated>2024-06-21T08:23:43.264Z</atom:updated>
            <content:encoded><![CDATA[<h4>From Local Integration test with Docker Compose to Scalable Cloud Apps with AWS Fargate and Terraform</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*283nLGXKnxJWYBs1" /></figure><p>Think about testing multiple microservices with their dependencies right in your local environment. Or running complex test cases that involve several interconnected services in your CI/CD pipeline. And then, deploying those same services to the cloud with minimal infrastructure management and operational overhead. Specially when you don’t want to manage Kubernetes Cluster. Recently, I received several inquiries on this very topic and decided it was time to put together a comprehensive demo project to address these challenges.</p><p>In this blog post, we’ll explore how to set up a local environment for testing integrated services and seamlessly deploy them to the cloud. The goal is to provide a quick and efficient way to handle development, testing, and deployment with minimal hassle. We’ll use Docker Compose for local testing and Terraform for cloud provisioning, ensuring a smooth transition from local development to production in AWS Fargate.</p><p>Let’s dive in and see how you can streamline your development workflow, making it easier to manage and deploy your microservices both locally and in the cloud.</p><h4>Project Overview</h4><p>In this project, we have a high-level architecture that consists of the following components:</p><ul><li><strong>Frontend Service</strong>: Built with Vue.js, the frontend service is responsible for the user interface. It makes API calls to the backend service to fetch data.</li><li><strong>Backend Service</strong>: Developed with Python Flask, the backend service handles API requests from the frontend. It processes these requests and interacts with the database to retrieve and return data.</li><li><strong>Database</strong>: A MySQL database that stores the list of books. The backend service queries this database to fetch the required data and serves it to the frontend.</li><li><strong>Tests: </strong>We have implemented tests to ensure the reliability and correctness of our application<br><strong><em>Backend Tests</em></strong>: Using Python’s requests library, we verify the functionality of the backend API endpoints, ensuring they process requests and interact with the database correctly.<br><strong><em>Frontend Tests</em></strong>: Using Playwright, we simulate user actions to validate that the frontend displays data fetched from the backend service correctly.</li></ul><p>The frontend service communicates with the backend service via HTTP requests. When the frontend needs to display a list of books, it sends a request to the backend API. The backend then queries the MySQL database, retrieves the list of books, and sends this data back to the frontend, which displays it to the user.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*16V_3u9ivK9h-2PTA6VOSw.png" /><figcaption>Cloud Deployment Architecture</figcaption></figure><p>Here you can find the source code for the project.</p><p><a href="https://github.com/lkravi/fargate-web-app">GitHub - lkravi/fargate-web-app</a></p><h4>Local Setup</h4><p>Setting up the local environment was the first step. Using Docker, we containerized our frontend, backend, and MySQL database to ensure consistency across development and production environments.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/327/1*C16LQZjHm6UJo1UzeyM8sQ.png" /></figure><p>Why Use Docker Compose?</p><p>Docker Compose is an excellent tool for defining and running multi-container applications. It allows us to quickly spin up a local integrated environment where all the services, including frontend, backend, and database, can run together seamlessly. This setup closely mirrors the production environment, making it easier to catch integration issues early in the development process. By using Docker Compose, we can ensure that our local environment is consistent and easily replicable, simplifying testing and debugging.</p><p>Docker Compose File:</p><pre>version: &#39;3&#39;<br>services:<br>  frontend:<br>    build: ./frontend<br>    ports:<br>      - &quot;8080:8080&quot;<br>    environment:<br>      - VUE_APP_BACKEND_URL=http://backend:5001<br>    depends_on:<br>      - backend<br>      - mysql<br><br>  backend:<br>    build: ./backend<br>    ports:<br>      - &quot;5001:5001&quot;<br>    environment:<br>      MYSQL_DATABASE_HOST: &quot;mysql&quot;<br>      MYSQL_DATABASE_PORT: &quot;3306&quot;<br>      MYSQL_DATABASE_USER: &quot;tf_rds_user1&quot;<br>      MYSQL_DATABASE_PASSWORD: &quot;password&quot;  # Set your password here<br>      MYSQL_DATABASE_DB: &quot;bookdb&quot;<br>      USE_LOCAL_DB: &quot;true&quot;<br>      AWS_REGION: &quot;us-east-1&quot;<br>      DB_MASTER_USER: &quot;root&quot;<br>      DB_MASTER_PASSWORD: &quot;rootpassword&quot;<br>    depends_on:<br>      - mysql<br><br>  mysql:<br>    image: mysql:8.0<br>    environment:<br>      MYSQL_ROOT_PASSWORD: &quot;rootpassword&quot;<br>    ports:<br>      - &quot;3306:3306&quot;<br>    volumes:<br>      - mysql-data:/var/lib/mysql<br>      - ./init.sql:/docker-entrypoint-initdb.d/init.sql<br>  <br>  test:<br>    build:<br>      context: ./tests<br>      dockerfile: Dockerfile.test<br>    depends_on:<br>      - frontend<br>      - backend<br>      - mysql<br><br>volumes:<br>  mysql-dat</pre><p>Running Locally</p><pre>docker-compose up --build</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zT_Oo187XwQdd_m2rSrExA.png" /></figure><p>Running the docker-compose up with build param will build your services and bring up the local env. You can access your frontend service on localhost:8080</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/887/1*QBXFMPkeV4ohXCtu1_Q3vg.png" /></figure><h4>Database Initialization and Data Seeding</h4><p>Local Database Initialization</p><p>For the local environment, we use a SQL script to initialize the database and seed it with some initial data. This script is executed automatically when the MySQL container starts, thanks to the Docker Compose setup. The init.sql file contains the necessary SQL commands to create the database schema and insert some sample data.</p><p>Cloud Database Initialization</p><p>For the cloud environment, we use an RDS database that resides in a private subnet without public access. Therefore, we cannot directly run SQL scripts on the RDS instance from our local machine. Instead, we use an AWS Lambda function to initialize the database and seed the data.</p><h4>Build Your Cloud Environment</h4><p>The Terraform script included in the repository is configured to set up all the necessary cloud resources. This includes the VPC network components, Security Groups, Load Balancers, Fargate Tasks, and Services. Simply run terraform apply, and it will provision all the required components and provide you with the frontend load balancer endpoint, which you can access through your browser.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F0Fw-oYZhPQM%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D0Fw-oYZhPQM&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F0Fw-oYZhPQM%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/edb4f4b0036d6f92e5c87fd09fab89d3/href">https://medium.com/media/edb4f4b0036d6f92e5c87fd09fab89d3/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=df262c78abcf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Mastering Service Mesh with Istio on AWS EKS using Terraform EKS Blueprints]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://lkravi.medium.com/mastering-service-mesh-with-istio-on-aws-eks-using-terraform-eks-blueprints-c1cab2c8e168?source=rss-4844d78af5------2"><img src="https://cdn-images-1.medium.com/max/1024/0*LXMIIM6EUjk12E9r" width="1024"></a></p><p class="medium-feed-snippet">Part 1</p><p class="medium-feed-link"><a href="https://lkravi.medium.com/mastering-service-mesh-with-istio-on-aws-eks-using-terraform-eks-blueprints-c1cab2c8e168?source=rss-4844d78af5------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://lkravi.medium.com/mastering-service-mesh-with-istio-on-aws-eks-using-terraform-eks-blueprints-c1cab2c8e168?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/c1cab2c8e168</guid>
            <category><![CDATA[istio]]></category>
            <category><![CDATA[service-mesh]]></category>
            <category><![CDATA[aws]]></category>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Wed, 27 Dec 2023 15:31:13 GMT</pubDate>
            <atom:updated>2023-12-28T00:30:18.570Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Terraform Vs OpenTofu]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://lkravi.medium.com/terraform-vs-opentofu-b4d2b519b3fa?source=rss-4844d78af5------2"><img src="https://cdn-images-1.medium.com/max/1144/1*JdeHWQ5P7rzEFvd6N09VuQ.png" width="1144"></a></p><p class="medium-feed-snippet">Recently, Hashicorp announced a license change for its products, including Terraform. It&#x2019;s been open source under an MPL v2 license for&#x2026;</p><p class="medium-feed-link"><a href="https://lkravi.medium.com/terraform-vs-opentofu-b4d2b519b3fa?source=rss-4844d78af5------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://lkravi.medium.com/terraform-vs-opentofu-b4d2b519b3fa?source=rss-4844d78af5------2</link>
            <guid isPermaLink="false">https://medium.com/p/b4d2b519b3fa</guid>
            <category><![CDATA[infrastructure-as-code]]></category>
            <category><![CDATA[iac]]></category>
            <category><![CDATA[opentofu]]></category>
            <category><![CDATA[terraform]]></category>
            <dc:creator><![CDATA[Ravi Yasakeerthi]]></dc:creator>
            <pubDate>Thu, 21 Sep 2023 01:46:36 GMT</pubDate>
            <atom:updated>2023-09-21T01:51:02.527Z</atom:updated>
        </item>
    </channel>
</rss>