Tableau Heartbeat Kept Databricks Warehouse Running All Weekend, Costing Team $14k

·5 views

A software team was hit with a $14,000 cloud bill after a newly deployed data pipeline triggered an unexpected cost spike over a single weekend on Databricks Serverless SQL Warehouse. The culprit turned out to be a Tableau BI tool configured to send periodic heartbeat pings to the warehouse every 8 minutes — just under the 10-minute Auto-stop idle threshold. Because Databricks Serverless treated these sub-millisecond system metadata queries as active sessions, the warehouse never registered as idle and never shut down. The team had unknowingly been running a large, high-DBU warehouse around the clock simply to serve low-cost connection keep-alive requests. The fix involved severing the BI connection, splitting workloads across smaller and larger warehouses by use case, and reconfiguring connection strings to prevent the same idle-timer reset from recurring.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

FoxyInvoice Dev Blog: How Invoice Lifecycle, Tax Engine, and Exports Were Built

Developer blog series FoxyInvoice has published Chapter 5, detailing the domain logic behind its invoicing platform built for real production use. The invoice lifecycle follows a state machine moving from Draft through Sent to Paid, Partial, Overdue, or Void, with overdue status derived daily rather than stored. A dedicated tax engine handles US sales tax complexity by evaluating exemptions, nexus, jurisdiction rates, and product taxability per line item, using decimal arithmetic to avoid floating-point errors. All totals are computed server-side, with the browser's figures never trusted, and invoices lock after a configurable window to prevent retroactive edits. The platform also supports export formats including CSV, QuickBooks IIF, and Tally XML to meet accountant requirements across different markets.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

9,560 MikroTik Routers Expose SSH Globally Amid Two Actively Exploited Flaws

CERT Polska disclosed six MikroTik RouterOS vulnerabilities in September 2026, two of which — CVE-2026-67277 and CVE-2026-86060 — were already being actively exploited in combination to give unauthenticated attackers full administrative control. CISA added both flaws to its Known Exploited Vulnerabilities catalog on 10 September 2026. A ZoomEye scan identified over 8 million internet-visible RouterOS devices, of which 9,560 specifically expose SSH — the key precondition for the reported attack chain. Security researchers caution that these figures represent a floor, not a ceiling, since scanning coverage is incomplete and the fingerprint cannot distinguish patched from unpatched firmware. Organizations running RouterOS equipment are advised to audit internet-facing SSH exposure, verify firmware versions against patched releases, and review logs for signs of unauthorized access.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Performance Tuning and Profiling Help Developers Build Faster Software

Performance tuning is a systematic process of optimizing software components — including code, databases, and networks — to improve speed, responsiveness, and resource efficiency. Profiling complements this by diagnosing runtime behavior, helping developers pinpoint CPU-heavy functions, memory leaks, and excessive I/O operations. The tuning cycle involves measuring a baseline, identifying bottlenecks, applying fixes, and re-measuring to confirm gains. Popular profiling tools include Chrome DevTools, Visual Studio Profiler, YourKit, Intel VTune, and Python's cProfile, each catering to different platforms and languages. Experts recommend targeting the most resource-intensive components first and controlling the profiling environment to ensure accurate, actionable results.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

AI Trading Systems Move Beyond Binary Vetoes With Conditional Approval Logic

A new architectural approach in AI-driven trading systems replaces binary pass-or-veto decisions with a 'Conditional Approval' mechanism that adjusts risk parameters dynamically. In a logged example from September 2026, an AI Advisor issued a PROCEED ruling on a UAIUSDT trade despite unfavorable macro conditions, applying tightened stop-losses and reduced position size instead of a blanket veto. The system detected an 88-score high-kinetic micro-structure signal indicating strong localized buying momentum, while simultaneously flagging elevated Bitcoin Dominance at 58.3% and broad market sell pressure as macro risks. Traditional binary logic would have forced a deadlock, either missing a profitable opportunity or exposing the portfolio to significant downside. The shift toward continuous risk space management allows AI trading systems to act as holistic risk managers rather than simple signal-to-order pipelines.

0 comments Read more at DEV Community