TCS DevOps Engineer Interview Questions

“I needed approval to upgrade the database version behind a client's loan application. I prepared a one-page plan: the reason for the upgrade, the exact steps, evidence from the rehearsal in the pre-production environment, the rollback steps with how long they take, and who would be on the call. The board asked what would happen if rollback failed, so I explained the backup restore we had also rehearsed. They moved the date away from month end, which was sensible. The upgrade went ahead in the next window. The result was your result.”

“An application server at a client ran out of disk space every few weeks, and each time someone deleted old logs and closed the ticket. I collected the dates and found it always followed a monthly report run. That job wrote large temporary files and did not clean up when it failed. I worked with the developers to fix the cleanup, moved temporary files to a separate volume, and added an alert that warned well before the disk filled. I raised it as a problem record so the history was linked. The result was your result.”

“I would first sit through one manual release to see where time and errors come from. Then I would pick one application with a willing team. The first step is an automated build and test on every commit, which changes nothing in production and shows value fast. Next, automated deployment to development and test environments using the same scripts each time. The client's approvals stay, but become gates in the pipeline with a record. Production deployment is automated last, run by the same people in the same window at first. Once releases are reliable, we can discuss doing them more often.”

“Environments drift when people fix things by hand in one place. I would define servers, settings and versions as code in version control, and build each environment from that. A scheduled job compares the real state with the definition and reports differences. The same build artifact moves through every stage, with only configuration changing. For test data, raw customer data should not leave production. I would use a masking process that replaces names, account numbers and contact details while keeping the shape of the data, or generate data where masking is not enough.”

“I would start by agreeing two targets with the client: how much data they can afford to lose and how long the service can be down. For online banking both are small. So the database is replicated to the second site continuously, and application servers there are kept built and patched from the same code as the main site. Switch-over steps are written as a runbook and automated where possible, including moving network traffic. One named person decides when to switch. The plan is only real if rehearsed, so I would schedule drills and run from the second site for a day.”

“The pipeline should refuse to deploy to production unless the build is linked to an approved change record and a reviewed merge. Test results, security scan results and the approver's name are attached to the build and stored where they cannot be edited. Nobody deploys to production by hand, so the pipeline log is the complete history. Each deployment records the version, the time, who approved it and what changed. With that in place, an auditor's question becomes a search, not a week of collecting emails. I would run a mock audit myself once a quarter to check that the records are complete.”

“Even though nothing is down, the client has had no working recovery for two weeks, and that is serious. I would first confirm the gap and how far behind the second site is. Then I would tell my lead and the client contact the same day, in plain words, with what I know and what I am doing. Next I would restore replication, watch it catch up, and run a check that the data at the second site matches. Then the harder question: why did nothing alert us? I would add a check on replication delay and review what else is quietly unmonitored.”

“I enjoy making releases dull, in a good way. At many client companies, especially banks and insurers, releases are still stressful nights, and the controls around them exist for real reasons. I find it rewarding to bring automation that keeps those controls and removes the manual errors. TCS works with clients across industries, so I would see many kinds of infrastructure, from old data centres to cloud. I have built pipelines and handled release nights, and I want to learn how to lead that change for a large client step by step.”