Social

Cum rezolvi problema Cold Start intr-o aplicatie sociala

Ai stat sase luni sa scrii cod la cea mai tare aplicatie, ai frontend in React, arhitectura serverless, design perfect. Urca aplicatia in productie. Intra un user real, isi face cont entuziasmat, ajunge pe feed-ul principal si vede un ecran complet albastru pe care scrie “No posts yet”. Nu urmareste pe nimeni, iar pe platforma bate vantul. Da uninstall sau inchide tab-ul fix in secunda doi. Bine ai venit in realitatea “Cold Start”.

[mai mult...]

Cum construiești un sistem „For You” când ai foarte puține date despre utilizator

Faci un proiect de aplicatie web care are un feed central de stiri, clipuri sau articole. Vrei neaparat sectiunea aia de “For You”, inspirata de TikTok. Dar te lovesti de o matematica brutala: nu ai un milion de utilizatori ca sa aplici algoritmi smecheri de tip Collaborative Filtering (adica sa ii recomanzi omului ce citesc alti oameni ca el). Utilizatorul tau abia si-a facut cont de 3 minute. Nu a dat niciun like, nu ai istoric, nu stii nimic despre el. Ce ii pui pe ecran?

[mai mult...]

Activarea funcției de recuperare/Recall al mesajelor între chiriași/tenant în Exchange Online

Funcția de retragere a mesajelor între tenants/chiriasi permite utilizatorilor să retragă e-mailurile trimise către o altă organizație Microsoft 365. Organizația destinatară trebuie să activeze această funcție și să adauge ID-ul chiriașului expeditorului pe lista sa de permisiuni. Funcția este dezactivată în mod implicit.

Exemplu: Clientul A trimite un e-mail către clientul B. Configurați clientul B astfel încât să permită retragerile de la clientul A. Pentru retragerile în ambele sensuri, configurați separat permisiunea inversă.

Cerințe preliminare

  1. Acces administrativ la tenantul destinatar.
  2. Un cont cu permisiunile RBAC pentru Exchange necesare pentru a rula comenzile de configurare.
  3. Modulul PowerShell pentru Exchange Online.
[mai mult...]

High availability vs. Disaster Recovery: designing enterprise infrastructure for real-world failures

High availability and disaster recovery are frequently discussed together, but they address different infrastructure challenges.

High availability focuses on maintaining service continuity despite the failure of individual components.Disaster recovery focuses on restoring services after disruptive events that exceed the protection provided by normal availability mechanisms.Confusing these objectives can lead to expensive infrastructure designs that still fail during major incidents.

High Availability Architecture

High availability typically relies on redundancy, health monitoring, automated failover, and mechanisms that reduce single points of failure.

Examples include clustered application servers, redundant network devices, load-balanced services, and database availability groups.However, redundant components do not automatically create a resilient system.Two servers connected to the same storage device may remain vulnerable to a shared storage failure.

Similarly, redundant applications depending on one authentication or network service may still experience a common outage.

Understanding Failure Domains

A failure domain represents a collection of components that can be affected by a shared event.Common failure domains include physical hosts, storage systems, network switches, power supplies, data centers, and cloud regions. Infrastructure architects should identify where redundancy actually exists and where dependencies remain shared.Availability design should be based on realistic failure scenarios rather than the number of deployed components.

Disaster recovery architectures may use backup restoration, storage replication, database replication, or standby environments.Synchronous replication can reduce data loss under suitable latency and connectivity conditions, but it may introduce performance and geographic constraints.

Asynchronous replication allows greater flexibility but generally introduces a potential gap between the latest committed production data and the replicated state.Replication must not be confused with backup.Accidental deletions, corrupt data, or malicious changes may also propagate to replicated environments.

Defining RPO and RTO

Recovery Point Objective establishes the maximum acceptable amount of data loss.Recovery Time Objective establishes how quickly business services must be restored.For example, a transaction-processing platform may require tighter recovery objectives than an internal reporting application.These requirements should be defined with business stakeholders and translated into technical architecture.

Failover and Failback Engineering

Failover transfers services to an alternative infrastructure environment.Failback returns operations to the original or newly restored primary environment.Both operations require documented procedures.DNS changes, application configurations, authentication dependencies, database consistency, and network routing must be considered.

Failback can be particularly challenging if production data has changed significantly while services operated in the recovery environment.

Testing Recovery Capability

A disaster recovery plan that has never been tested provides limited assurance.Exercises should verify the complete recovery sequence, including infrastructure startup, application dependencies, data consistency, and user access.

Testing results should be measured against defined recovery objectives.

[mai mult...]

Advanced DNS troubleshooting in enterprise Networks: resolution Paths, caching, and Split-Horizon architecture

DNS is one of the most fundamental dependencies in enterprise infrastructure.

Authentication, application connectivity, service discovery, email delivery, and hybrid cloud integration frequently depend on reliable name resolution.DNS failures can produce symptoms that closely resemble routing problems, application outages, or authentication failures.Consequently, effective troubleshooting requires understanding the complete resolution path rather than simply checking whether a hostname resolves on one workstation.

When a client requests a DNS record, the response may originate from a local cache, recursive resolver, or authoritative DNS server.Caching improves efficiency but can complicate incident investigation.

An outdated cached record may continue directing clients toward an unavailable endpoint until its time-to-live expires or the cached entry is invalidated.Engineers should distinguish between authoritative data, resolver behavior, and client-side caching.

Split-Horizon DNS

Split-horizon DNS allows different DNS responses to be provided according to the resolution environment or policy.

For example, an internal enterprise application may resolve to a private address for corporate clients while external users receive a public endpoint.This architecture can improve network control but introduces operational complexity. Incorrectly configured internal zones may prevent clients from resolving records that are valid in public DNS.

Active Directory DNS Dependencies

Active Directory environments rely heavily on DNS records for locating domain controllers and associated directory services.Service records enable clients to discover suitable authentication and directory endpoints.

Incorrect DNS client configurations, unavailable domain controller records, or replication inconsistencies may result in authentication problems.Administrators should validate both name resolution and the availability of the services referenced by DNS.

Conditional forwarders can direct queries for selected namespaces to designated DNS servers.They are particularly useful in environments integrating multiple internal namespaces or cloud-hosted private services.However, forwarder loops, unreachable resolvers, and inconsistent zone ownership can create difficult-to-diagnose failures.

DNS architecture should include documented ownership, redundancy, and resolution paths.

Effective DNS investigations should establish:

  • Which resolver the affected client uses.
  • Whether the requested record exists authoritatively.
  • Whether cached data differs from authoritative results.
  • Whether the expected namespace is resolved through forwarding.
  • Whether the returned addresses are reachable.
  • Whether the application requires additional service records.

Packet captures and DNS server logs may be necessary when ordinary lookup tools fail to explain inconsistent behavior.

[mai mult...]