ranjan@ranjan.info:~$ git log --grep="resolved"

Problems that got solved

Real engagements, written up without the confidential parts — the problem as it arrived, the approach that fixed it, and what the client measured afterwards. Names withheld; outcomes verifiable on request.

Emergency Recovery

Production server recovered after a failed kernel update

The problem
An e-commerce company's primary server kernel-panicked after an unattended upgrade — no boot, no recent snapshot, peak sales week.
What I did
Booted rescue mode over the provider console, repaired the boot chain against a known-good kernel, verified filesystem integrity, and documented the failed package hold that caused it.
The result
Back online in under 4 hours with zero data loss. Unattended upgrades replaced with staged, monitored patching.
  • < 4h to recovery
  • 0 bytes lost
  • Root cause documented
GRUBext4Rescue modeIPMI
Migration

Hundreds of websites moved between data centers without downtime

The problem
A hosting reseller had to leave a deteriorating provider — 400+ cPanel accounts, live mail, and customers who would notice any outage.
What I did
Staged full-account syncs to the new fleet, verified sites and mailboxes account-by-account, lowered TTLs days ahead, then executed a scheduled overnight DNS cutover with the old fleet held for rollback.
The result
All accounts live on new infrastructure by morning. Zero support tickets attributable to the move.
  • 400+ accounts
  • 0 downtime tickets
  • 1 overnight cutover
cPanel · WHMrsyncimapsyncCloudflare DNS
Malware Removal

Malware cleaned from a multi-site portfolio, blacklists cleared

The problem
An agency's shared server was reinfected three times after cleanups by generalist support — Google had blacklisted key client sites and mail was bouncing.
What I did
Correlated file modification times against access logs to find the actual entry point (an abandoned plugin), removed layered backdoors across accounts, isolated sites from each other, and submitted delisting requests.
The result
Clean for good — no reinfection since. All blacklistings cleared within days; deliverability restored.
  • 12 sites cleaned
  • 0 reinfections since
  • Blacklists cleared
WordPressmaldetModSecuritySearch Console
Performance

Overloaded Linux infrastructure tuned instead of upgraded

The problem
A SaaS on a dedicated server hit constant load spikes; the vendor recommendation was doubling hardware spend.
What I did
Profiled the stack under real traffic: found unindexed hot queries, an undersized InnoDB buffer pool, no object cache, and PHP-FPM pools sized for a different era. Fixed each layer with before/after benchmarks.
The result
Page times dropped by more than half and load spikes disappeared — on the same hardware. The upgrade was cancelled.
  • ~60% faster pages
  • $0 new hardware
  • Load spikes eliminated
MySQLRedisPHP-FPMNginx
Automation

Hosting operations automated end to end

The problem
A growing host was drowning in manual work: provisioning by checklist, backups checked "when someone remembered", suspensions done by hand.
What I did
Built WHMCS-driven provisioning hooks, automated backup verification with alerting on failure (not just on error logs nobody read), and a small ops dashboard showing fleet state at a glance.
The result
Provisioning went from 30 minutes to instant; backup failures now page within minutes; the founder got out of the ops loop.
  • 30 min → instant provisioning
  • 100% verified backups
  • 1 dashboard
WHMCSPHPBashJetBackup
Architecture

Scalable hosting architecture designed for a product launch

The problem
A company launching a hosted product needed infrastructure that could start cheap and scale without a rebuild — and had been quoted an enterprise-sized bill.
What I did
Designed a tiered architecture on commodity cloud: CloudLinux-isolated app tier, replicated database, object storage for user data, and load balancing that could grow from one node to many with configuration, not surgery.
The result
Launched at a fraction of the quoted cost, and scaled through growth spikes with node additions instead of migrations.
  • ~70% below quoted cost
  • Scale-out, not rebuild
  • Launch on time
CloudLinuxMariaDB replicationNginx LBHetzner
ranjan@ranjan.info:~$ cp your-problem ~/inbox/

Your situation could be the next one

Every engagement above started with a short email describing a problem. Send yours — worst case, you get a free straight answer about what it would take.

Start a conversationView services