SShortSingh.
Back to feed

FrankenPHP Worker Mode Cuts Tenant-Switch Cost in Shared Symfony 8.1 Deployments

0
·2 views

A follow-up to an earlier article examines running a database-per-tenant SaaS on a shared Symfony 8.1 deployment with FrankenPHP's worker mode. The author notes that under php-fpm each request re-boots the kernel, re-queries Redis for tenant placement, and opens a fresh database connection, costing about 5.8 ms per tenant switch. FrankenPHP keeps the kernel alive across requests, which can eliminate much of that per-request overhead. The article also warns that worker mode changes which state persists between requests, so care is needed to avoid leaking data between tenants. The author plans to compare four deployments under a load that switches tenants on every request. The demo uses PHP 8.4, PostgreSQL 16, Redis 7 and FrankenPHP 1.13, and the Symfony runtime package is used because the separate FrankenPHP adapter does not yet support Symfony 8.1.

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)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

How to Diagnose Slow Page Interactions Using INP and Three Latency Phases

A guide explains why a page can load quickly yet still feel sluggish when users click, type, or submit. It introduces Interaction to Next Paint (INP), a Core Web Vital that measures responsiveness across a visit, and outlines Google's thresholds of 200 ms or less for good and above 500 ms for poor at the 75th percentile. The author argues that slow interactions break into three phases: input delay, processing duration, and presentation delay, each with different causes and fixes. Typical culprits include busy main threads, heavy JavaScript handlers, and rendering work after scripts finish. The article recommends using Chrome DevTools, real-user data, and Google's web-vitals library to locate the actual bottleneck before optimizing.

0
ProgrammingDEV Community ·

TranscribeLink extracts audio in the browser so large videos upload faster

Transcribing long screen recordings usually means uploading multi-gigabyte video files, even though speech models only need the audio. The TranscribeLink developer addresses this by having the browser extract the audio track and upload just that, which shrinks an hour of speech to roughly 14 MB. The extraction uses WebCodecs together with the Mediabunny library, which handles container formats and re-encodes audio to Opus or AAC depending on browser support. Extraction runs in parallel with the original upload, and the smaller audio file replaces the video when it finishes first, with the original uploaded as a fallback.

0
ProgrammingDEV Community ·

Chatwoot's licensing: open-source core, enterprise directory under separate terms

A developer reviewed Chatwoot's licenses, pricing and code, arguing that the popular label of a free Intercom alternative obscures how the offering is actually structured. The root repository uses the MIT Expat license, while the enterprise/ directory falls under its own license requiring a paid subscription with a seat count for production use. Cloud pricing ranges from a free tier with limited agents to $99 per agent per month for the enterprise plan, which adds SSO and audit logs. The Captain AI feature is a managed, credit-based product not included in the free edition, though the vendor points to Agent Bots and the API as ways to connect a self-chosen model and pay only for tokens. The author concludes the trade-off suits teams wanting self-hosted data and multichannel inboxes, and that developers embedding support will likely find the API sufficient.

0
ProgrammingDEV Community ·

Study Ranks Which Software Products Draw the Most Referral Traffic from AI Engines

A November 2026 snapshot from NeewP's GEO AI Traffic Report analyzed 535 growth-tracked products, of which 114 showed measurable referral traffic from generative AI platforms. ChatGPT appeared in 110 of those 114 products, with Gemini close behind and Claude strong among developer and workflow tools. Creative AI tools, AI research search, coding tools, and developer utilities captured the largest shares of their site traffic from AI engines, with smaller category-defining tools like tldraw.com outperforming many larger SaaS suites. The authors caution that the data relies on Semrush and AICPB attribution signals, which understate zero-click mentions and dark referrals, and that the dataset favors high-visibility domains.