{"id":76,"date":"2025-05-13T10:23:00","date_gmt":"2025-05-13T10:23:00","guid":{"rendered":"https:\/\/wordpress.joeltan.me\/?p=76"},"modified":"2026-01-26T10:34:58","modified_gmt":"2026-01-26T10:34:58","slug":"my-data-engineering-wake-up-call-from-3-am-alerts-to-streaming-serenity","status":"publish","type":"post","link":"https:\/\/joeltan.me\/?p=76","title":{"rendered":"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity"},"content":{"rendered":"\n

Data pipelines are the unsung heroes of the analytics world. When they work, insights flow; when they falter, the entire data-driven engine can grind to a halt. I’ve always appreciated a well-designed data model, but experience has taught me a crucial lesson: without fresh, reliable data, even the most elegant front-end visualizations are just window dressing.<\/p>\n\n\n\n

I was responsible for a traditional data warehouse, tasked with delivering timely insights. The reality, however, was a constant battle with delta load processes. We were mired in control tables, complex SQL, and the ever-present anxiety of late-night alerts. If this sounds familiar, this story is for you.<\/p>\n\n\n\n

My Old Life: A Daily Battle I Was Losing<\/h2>\n\n\n\n

Let me paint you a picture of what my typical day looked like before the transformation. I’d arrive at the office, coffee in hand, and immediately check our monitoring dashboard with the same anxiety most people reserve for checking their bank balance after an overseas holidays.<\/p>\n\n\n\n

Control Tables: My Personal Hell<\/h2>\n\n\n\n

I built this elaborate control table system. One master table tracking “last processed timestamp” for every data source. Seemed smart at the time.<\/p>\n\n\n\n

It wasn’t.<\/p>\n\n\n\n

I spent 40% of my time just keeping this thing alive. New data source? New row. Edge case? New column. Corruption? New migraine.<\/p>\n\n\n\n

The table had 200+ rows. Each one a potential point of failure. When it went down, everything went down. And it went down a lot.<\/p>\n\n\n\n

SQL Scripts That Haunted Me<\/h2>\n\n\n\n

I wrote some truly horrific SQL. My “masterpiece” was a 200-line query for order summaries. I actually named it “The Beast.”<\/p>\n\n\n

\n-- This thing made grown DBAs cry\nSELECT\n    o.order_id,\n    c.customer_name,\n    CASE\n        WHEN t.order_id IS NULL THEN 'INSERT'\n        WHEN MD5(CONCAT(o.order_amount, c.customer_name)) <> t.row_hash THEN 'UPDATE'\n        ELSE 'NO_CHANGE'\n    END as change_type\nFROM orders o\nJOIN customers c ON o.customer_id = c.customer_id\nLEFT JOIN target_orders_summary t ON o.order_id = t.order_id\nWHERE o.last_updated_ts > (SELECT last_processed_ts FROM control_table WHERE source_name = 'orders')\n<\/pre><\/div>\n\n\n
MD5 hashes for change detection. Seemed clever until someone added a column and forgot to update the hash logic. Suddenly every row looked “unchanged” while being completely different.<\/p>\n\n\n\n
Took me three days to figure out why the new commission field wasn’t showing up.<\/p>\n\n\n\n
My Personal “Nuclear Option” Addiction<\/h2>\n\n\n\n
When things went wrong (and they often did), I had developed what I affectionately called my “nuclear option” – full table reloads. Source system acting up? Nuclear option. Timestamps looking funky? Nuclear option. Tuesday? Might as well go nuclear.<\/p>\n\n\n\n
I justified this by telling myself it was “ensuring data integrity,” but really, I was just scared of missing something. The problem was that these full reloads were killing our source systems and blowing out our processing windows. I once crashed our CRM system because I was pulling the entire customer table for the third time that week.<\/p>\n\n\n\n
The Day Everything Changed<\/h2>\n\n\n\n
Tuesday, March 15th. Our main source system added milliseconds to their timestamps. Went from YYYY-MM-DD HH:MM:SS to YYYY-MM-DD HH:MM:SS.fff.<\/p>\n\n\n\n
Tiny change. Massive problem.<\/p>\n\n\n\n
My parsing logic broke. The system thought every record was new. All million of them. Processing loop from hell.<\/p>\n\n\n\n
Eighteen hours of downtime. Post-incident meeting where everyone realized how fragile our setup really was.<\/p>\n\n\n\n
That night, 2 AM, staring at error logs, I decided: never again.<\/p>\n\n\n\n
Research Phase: Finding a Better Way<\/h2>\n\n\n\n
I spent weeks reading everything about modern data architecture. Delta Lake kept coming up. Specifically, Change Data Feed (CDF).<\/p>\n\n\n\n
The promise: automatic change tracking. No custom SQL nightmares.<\/p>\n\n\n\n
I was skeptical. After years of controlling everything, trusting a platform felt wrong.<\/p>\n\n\n\n
But I was exhausted.<\/p>\n\n\n\n
First Test: Mind Blown<\/h2>\n\n\n\n
Picked my worst data source – a customer table that had tortured me for two years. Inconsistent timestamps, random schema changes, surprise duplicates.<\/p>\n\n\n\n
New approach:<\/p>\n\n\n
\n# Enable CDF (one time)\nspark.sql("ALTER TABLE customer_delta SET TBLPROPERTIES (delta.enableChangeDataFeed = true)")\n\n# Get changes\nchanges_df = spark.read.format("delta") \\\n    .option("readChangeFeed", "true") \\\n    .option("startingVersion", 10) \\\n    .option("endingVersion", 15) \\\n    .table("customer_delta")\n\nchanges_df.show()\n<\/pre><\/div>\n\n\n
Three hours of work became fifteen minutes. 200 lines of SQL became 10 lines of Python.<\/p>\n\n\n\n
I ran it five times to make sure it wasn’t a fluke.<\/p>\n\n\n\n
Building Our New Data Superhighway<\/h2>\n\n\n\n
With the PoC validating our approach, we moved to implement a new architecture based on the medallion model, a layered approach to progressively refine data:<\/p>\n\n\n\n
    \n
  • Bronze Layer:<\/strong> This is where raw, untouched data lands, typically in simple blob storage. It serves as our pristine archive, preserving data in its original state.<\/li>\n\n\n\n
  • Silver Layer:<\/strong> Here, data from the Bronze layer is consolidated, cleansed, and, most importantly, transformed into the Delta Lake format with CDF enabled<\/em>. This layer became our new single source of truth for “what changed.”<\/li>\n\n\n\n
  • Gold Layer:<\/strong> This layer houses business-ready, aggregated, and curated datasets, also in Delta Lake format. It provides optimized data for analytics and reporting, complete with version history and time-travel capabilities (which are as powerful as they sound!).<\/li>\n<\/ul>\n\n\n\n
    Alongside the medallion architecture, we also prioritized:<\/p>\n\n\n\n
      \n
    • True Change Data Capture (CDC):<\/strong> Wherever possible, we implemented direct CDC by tapping into source database transaction logs, ensuring the most accurate and low-latency capture of changes.<\/li>\n\n\n\n
    • Near Real-Time Streaming:<\/strong> For critical datasets, we transitioned from batch processing to micro-batch or full streaming pipelines, leveraging Delta Lake’s capabilities as both a streaming source and sink. This dramatically reduced data latency.<\/li>\n<\/ul>\n\n\n
      \n    # Conceptual PySpark Structured Streaming from a Delta table (ideally with CDF)\n    # Illustrating a stream-to-stream join\n\n    # Stream 1: Orders (assuming CDF is enabled)\n    orders_stream_df = spark.readStream.format("delta") \\\n        .option("readChangeData", "true") \\\n        .option("startingVersion", "latest") \\\n        .table("delta_orders_table_with_cdf") \\\n        .selectExpr("order_id", "customer_id AS o_customer_id", "order_amount", "order_timestamp", "_change_type AS order_change_type")\n\n    # Stream 2: Customers (assuming CDF is enabled)\n    customers_stream_df = spark.readStream.format("delta") \\\n        .option("readChangeData", "true") \\\n        .option("startingVersion", "latest") \\\n        .table("delta_customers_table_with_cdf") \\\n        .selectExpr("customer_id", "customer_name", "customer_address", "_change_type AS customer_change_type")\n\n    # Watermarking for stateful operations like joins\n    orders_with_watermark = orders_stream_df \\\n        .withWatermark("order_timestamp", "10 minutes") # Allow data to be 10 mins late\n\n    customers_with_watermark = customers_stream_df \\\n        .withWatermark("event_timestamp", "10 minutes") # Assuming customers have an event_timestamp for changes\n\n    # Stream-to-stream join\n    # This requires watermarks on both sides and an event-time constraint\n    joined_stream_df = orders_with_watermark.join(\n        customers_with_watermark,\n        expr("""\n            o_customer_id = customer_id AND\n            order_timestamp >= event_timestamp - interval 1 hour AND\n            order_timestamp <= event_timestamp + interval 1 hour\n            """),\n        "inner" # or leftOuter, rightOuter, etc.\n    )\n\n    # Define some transformation for the joined stream\n    def process_joined_data(batch_df, batch_id):\n        # Now batch_df contains joined data from orders and customers\n        # You can apply further logic here, e.g., update an aggregate table\n        print(f"Processing joined batch {batch_id}, count: {batch_df.count()}")\n        if not batch_df.isEmpty():\n            # Example: Write to a consolidated Delta table\n            batch_df.write.format("delta").mode("append").saveAsTable("joined_orders_customers_summary")\n\n    # Write the joined stream to a sink\n    query = joined_stream_df.writeStream \\\n        .foreachBatch(process_joined_data) \\\n        .outputMode("append") \\ # Append is common for stream-stream joins\n        .option("checkpointLocation", "\/path\/to\/joined_checkpoint\/dir") \\\n        .trigger(processingTime="1 minute") \\\n        .start() # Use .start() and then manage the query object\n\n    # query.awaitTermination() # In a real script\n<\/pre><\/div>\n\n\n
      Surprisingly, migrating many of our old pipelines proved less painful than anticipated, as the clarity and built-in capabilities of the new platform simplified much of the logic.<\/p>\n\n\n\n
      The Learning Curve: My Personal Struggles and Breakthroughs<\/h2>\n\n\n\n
      The transition wasn’t without its challenges. I had to completely rewire my brain from thinking like a database developer to thinking like a platform engineer.<\/p>\n\n\n\n
      Performance Optimization<\/strong>: I learned the hard way that Delta Lake performance requires understanding partitioning and Z-ordering. My first attempts were embarrassingly slow until I discovered these commands:<\/p>\n\n\n
      \nOPTIMIZE my_table ZORDER BY (customer_id, order_date);\n<\/pre><\/div>\n\n\n
      Managing History<\/strong>: The transaction logs grew faster than I expected. I had to learn about VACUUM and retention policies. Finding the sweet spot between performance and audit requirements took some trial and error (and a few tense conversations with our compliance team).<\/p>\n\n\n\n
      Letting Go of Control<\/strong>: This was the hardest part for me personally. I had to stop trying to micro-manage every aspect of change detection and trust the platform to do its job. It felt like learning to drive all over again.<\/p>\n\n\n\n
      Results<\/h2>\n\n\n\n
      Six months later:<\/p>\n\n\n\n
      87% faster processing
      65% lower costs
      Zero 3 AM alerts<\/p>\n\n\n\n
      But the real win: I don’t dread Monday mornings anymore.<\/p>\n\n\n\n
      What I’d Tell Past Me<\/h2>\n\n\n\n
        \n
      1. Fix your biggest pain point first<\/li>\n\n\n\n
      2. Trust the platform – your custom solutions aren’t that clever<\/li>\n\n\n\n
      3. Document everything<\/li>\n\n\n\n
      4. It’s okay to let go of control<\/li>\n<\/ol>\n\n\n\n
        Still Learning<\/h2>\n\n\n\n
        The modernization changed more than just my technical architecture – it changed my entire relationship with data engineering. I went from reactive maintenance to proactive innovation. Instead of being the person who fixes broken pipelines, I became the person who builds platforms that don’t break.<\/p>\n\n\n\n
        Bottom Line<\/h2>\n\n\n\n
        That Tuesday crisis was the best thing that happened to my career. It forced me to stop patching and start fresh.<\/p>\n\n\n\n
        If you’re debugging at 2 AM, tired of explaining why data is always late, constantly fixing instead of building – there’s a better way.<\/p>\n\n\n\n
        Sometimes you have to admit your approach isn’t working and start over.<\/p>\n\n\n\n
        Your future self will thank you. Mine does every night when I actually sleep..<\/p>\n","protected":false},"excerpt":{"rendered":"
        Data pipelines are the unsung heroes of the analytics world. When they work, insights flow; when they falter, the entire...<\/p>\n","protected":false},"author":1,"featured_media":77,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_import_markdown_pro_load_document_selector":0,"_import_markdown_pro_submit_text_textarea":"","footnotes":""},"categories":[3],"tags":[],"class_list":["post-76","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-data-engineering-systems-that-stay-up"],"yoast_head":"\nMy Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity - Joel Tan Tech Blogs<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/joeltan.me\/?p=76\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity - Joel Tan Tech Blogs\" \/>\n<meta property=\"og:description\" content=\"Data pipelines are the unsung heroes of the analytics world. When they work, insights flow; when they falter, the entire...\" \/>\n<meta property=\"og:url\" content=\"https:\/\/joeltan.me\/?p=76\" \/>\n<meta property=\"og:site_name\" content=\"Joel Tan Tech Blogs\" \/>\n<meta property=\"article:published_time\" content=\"2025-05-13T10:23:00+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-01-26T10:34:58+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"2560\" \/>\n\t<meta property=\"og:image:height\" content=\"1674\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Joel Tan\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Joel Tan\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/joeltan.me\/?p=76#article\",\"isPartOf\":{\"@id\":\"https:\/\/joeltan.me\/?p=76\"},\"author\":{\"name\":\"Joel Tan\",\"@id\":\"https:\/\/joeltan.me\/#\/schema\/person\/db13342201787db723bfdeadcd792743\"},\"headline\":\"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity\",\"datePublished\":\"2025-05-13T10:23:00+00:00\",\"dateModified\":\"2026-01-26T10:34:58+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/joeltan.me\/?p=76\"},\"wordCount\":1149,\"commentCount\":0,\"image\":{\"@id\":\"https:\/\/joeltan.me\/?p=76#primaryimage\"},\"thumbnailUrl\":\"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg\",\"articleSection\":[\"Data Engineering: Systems That Stay Up\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/joeltan.me\/?p=76#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/joeltan.me\/?p=76\",\"url\":\"https:\/\/joeltan.me\/?p=76\",\"name\":\"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity - Joel Tan Tech Blogs\",\"isPartOf\":{\"@id\":\"https:\/\/joeltan.me\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/joeltan.me\/?p=76#primaryimage\"},\"image\":{\"@id\":\"https:\/\/joeltan.me\/?p=76#primaryimage\"},\"thumbnailUrl\":\"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg\",\"datePublished\":\"2025-05-13T10:23:00+00:00\",\"dateModified\":\"2026-01-26T10:34:58+00:00\",\"author\":{\"@id\":\"https:\/\/joeltan.me\/#\/schema\/person\/db13342201787db723bfdeadcd792743\"},\"breadcrumb\":{\"@id\":\"https:\/\/joeltan.me\/?p=76#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/joeltan.me\/?p=76\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/joeltan.me\/?p=76#primaryimage\",\"url\":\"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg\",\"contentUrl\":\"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg\",\"width\":2560,\"height\":1674,\"caption\":\"A speed hump warning sign under a cloudy sky, emphasizing safety.\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/joeltan.me\/?p=76#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/joeltan.me\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/joeltan.me\/#website\",\"url\":\"https:\/\/joeltan.me\/\",\"name\":\"Joel Tan Tech Blogs\",\"description\":\"Building systems that survive real life\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/joeltan.me\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/joeltan.me\/#\/schema\/person\/db13342201787db723bfdeadcd792743\",\"name\":\"Joel Tan\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/joeltan.me\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/secure.gravatar.com\/avatar\/d9b5d1ab218cb2478280027d371ea60543f6551132d31a8cbd45a5a5b3fbadc9?s=96&d=mm&r=g\",\"contentUrl\":\"https:\/\/secure.gravatar.com\/avatar\/d9b5d1ab218cb2478280027d371ea60543f6551132d31a8cbd45a5a5b3fbadc9?s=96&d=mm&r=g\",\"caption\":\"Joel Tan\"},\"sameAs\":[\"http:\/\/192.168.1.146\"],\"url\":\"https:\/\/joeltan.me\/?author=1\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity - Joel Tan Tech Blogs","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/joeltan.me\/?p=76","og_locale":"en_US","og_type":"article","og_title":"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity - Joel Tan Tech Blogs","og_description":"Data pipelines are the unsung heroes of the analytics world. When they work, insights flow; when they falter, the entire...","og_url":"https:\/\/joeltan.me\/?p=76","og_site_name":"Joel Tan Tech Blogs","article_published_time":"2025-05-13T10:23:00+00:00","article_modified_time":"2026-01-26T10:34:58+00:00","og_image":[{"width":2560,"height":1674,"url":"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg","type":"image\/jpeg"}],"author":"Joel Tan","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Joel Tan","Est. reading time":"8 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/joeltan.me\/?p=76#article","isPartOf":{"@id":"https:\/\/joeltan.me\/?p=76"},"author":{"name":"Joel Tan","@id":"https:\/\/joeltan.me\/#\/schema\/person\/db13342201787db723bfdeadcd792743"},"headline":"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity","datePublished":"2025-05-13T10:23:00+00:00","dateModified":"2026-01-26T10:34:58+00:00","mainEntityOfPage":{"@id":"https:\/\/joeltan.me\/?p=76"},"wordCount":1149,"commentCount":0,"image":{"@id":"https:\/\/joeltan.me\/?p=76#primaryimage"},"thumbnailUrl":"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg","articleSection":["Data Engineering: Systems That Stay Up"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/joeltan.me\/?p=76#respond"]}]},{"@type":"WebPage","@id":"https:\/\/joeltan.me\/?p=76","url":"https:\/\/joeltan.me\/?p=76","name":"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity - Joel Tan Tech Blogs","isPartOf":{"@id":"https:\/\/joeltan.me\/#website"},"primaryImageOfPage":{"@id":"https:\/\/joeltan.me\/?p=76#primaryimage"},"image":{"@id":"https:\/\/joeltan.me\/?p=76#primaryimage"},"thumbnailUrl":"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg","datePublished":"2025-05-13T10:23:00+00:00","dateModified":"2026-01-26T10:34:58+00:00","author":{"@id":"https:\/\/joeltan.me\/#\/schema\/person\/db13342201787db723bfdeadcd792743"},"breadcrumb":{"@id":"https:\/\/joeltan.me\/?p=76#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/joeltan.me\/?p=76"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/joeltan.me\/?p=76#primaryimage","url":"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg","contentUrl":"https:\/\/joeltan.me\/wp-content\/uploads\/2026\/01\/a-speed-hump-warning-sign-under-a-cloudy-sky-emphasizing-safety.-6796809-scaled.jpg","width":2560,"height":1674,"caption":"A speed hump warning sign under a cloudy sky, emphasizing safety."},{"@type":"BreadcrumbList","@id":"https:\/\/joeltan.me\/?p=76#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/joeltan.me\/"},{"@type":"ListItem","position":2,"name":"My Data Engineering Wake-Up Call: From Azure Alerts to Streaming Serenity"}]},{"@type":"WebSite","@id":"https:\/\/joeltan.me\/#website","url":"https:\/\/joeltan.me\/","name":"Joel Tan Tech Blogs","description":"Building systems that survive real life","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/joeltan.me\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/joeltan.me\/#\/schema\/person\/db13342201787db723bfdeadcd792743","name":"Joel Tan","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/joeltan.me\/#\/schema\/person\/image\/","url":"https:\/\/secure.gravatar.com\/avatar\/d9b5d1ab218cb2478280027d371ea60543f6551132d31a8cbd45a5a5b3fbadc9?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/d9b5d1ab218cb2478280027d371ea60543f6551132d31a8cbd45a5a5b3fbadc9?s=96&d=mm&r=g","caption":"Joel Tan"},"sameAs":["http:\/\/192.168.1.146"],"url":"https:\/\/joeltan.me\/?author=1"}]}},"_links":{"self":[{"href":"https:\/\/joeltan.me\/index.php?rest_route=\/wp\/v2\/posts\/76","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/joeltan.me\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/joeltan.me\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/joeltan.me\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/joeltan.me\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=76"}],"version-history":[{"count":2,"href":"https:\/\/joeltan.me\/index.php?rest_route=\/wp\/v2\/posts\/76\/revisions"}],"predecessor-version":[{"id":80,"href":"https:\/\/joeltan.me\/index.php?rest_route=\/wp\/v2\/posts\/76\/revisions\/80"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/joeltan.me\/index.php?rest_route=\/wp\/v2\/media\/77"}],"wp:attachment":[{"href":"https:\/\/joeltan.me\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=76"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/joeltan.me\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=76"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/joeltan.me\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=76"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}