HireHireInterview Quizzes › Support Engineer

Support Engineer Interview Questions

Think you're ready? These are the questions that actually decide Support Engineer interviews. Warm up on Easy — then face the Hard round, where 95% of candidates crumble. 80 questions across 3 levels, instant score, completely free.

80Questions
3Difficulty levels
95%Fail the hard round
FreeInstant score
Easy
Warm-up · 20 Qs
Medium
Practical · 30 Qs
Hard
Brutal · 30 Qs
⚡ Take the Support Engineer quiz — get your score →

The Support Engineer interview questions

Below are the real questions, grouped by difficulty. Expand any one to reveal the correct answer and why — or take the timed quiz for a score you can share. Can you clear the Hard round?

Easy round 20 questions

A customer reports a web page shows "500 Internal Server Error". Where should you look FIRST to diagnose it?
  • A. The user's browser cache
  • B. The user's internet speed
  • C. The server-side application logs ✓
  • D. The user's DNS settings
Correct answer: C. A 500 is a server-side failure, so the application/server logs are the first place to investigate.
An API returns "403 Forbidden" even though the user logged in successfully. What does this most likely indicate?
  • A. The user is authenticated but lacks permission for that resource ✓
  • B. The user's password is wrong
  • C. The server is completely down
  • D. The requested URL does not exist
Correct answer: A. 403 means the request was understood and the identity known, but the user isn't authorized for that action.
You can successfully ping a server's IP, but its website won't load in a browser. What is the most likely cause?
  • A. The network cable is unplugged
  • B. DNS is completely broken
  • C. The server has no IP address
  • D. The web service or its port is down while the host is up ✓
Correct answer: D. A successful ping proves the host is reachable, so the problem is the web service/port, not connectivity.
A P1 ticket has a 4-hour resolution SLA, was logged at 10:00 AM, and is still unresolved at 1:30 PM. What is the best action?
  • A. Wait until 2:00 PM, then start acting
  • B. Proactively escalate before the SLA breaches ✓
  • C. Close the ticket as resolved
  • D. Reassign it to another queue silently
Correct answer: B. Good practice is to escalate before a breach, not after, to protect the SLA and the customer.
A user can reach a site by typing its IP address but not by its domain name. What is the likely cause?
  • A. The web server is down
  • B. The SSL certificate expired
  • C. A DNS resolution problem ✓
  • D. A blocked firewall port
Correct answer: C. If the IP works but the name doesn't, name-to-IP resolution (DNS) is failing.
You need to watch the newest lines of a log file as they are written in real time. Which command fits best?
  • A. cat app.log
  • B. head app.log
  • C. grep app.log
  • D. tail -f app.log ✓
Correct answer: D. tail -f streams new lines as they are appended, ideal for live log monitoring.
Multiple users in different cities report the same application is down at the same time. What does this most strongly suggest?
  • A. A server-side or backend outage ✓
  • B. A local issue on one user's machine
  • C. A single browser plugin conflict
  • D. One expired user password
Correct answer: A. A shared failure across many independent users points to a central server-side problem.
A customer reports a bug you cannot reproduce on your end. What is the best next step?
  • A. Close it as invalid
  • B. Gather exact steps, screenshots, and environment details ✓
  • C. Tell them to reinstall their operating system
  • D. Escalate to engineering immediately
Correct answer: B. Collecting reproduction steps and environment info is the standard way to make an issue reproducible.
Users start getting "502 Bad Gateway" errors. What does this typically indicate?
  • A. The client sent a malformed request
  • B. The requested page was not found
  • C. An upstream backend returned an invalid response to the gateway ✓
  • D. The user's authentication token expired
Correct answer: C. 502 means the gateway/proxy got a bad response from the upstream server it depends on.
You're asked to find how many rows are in the orders table without changing any data. Which query is safe to run?
  • A. DELETE FROM orders
  • B. UPDATE orders SET total = 0
  • C. DROP TABLE orders
  • D. SELECT COUNT(*) FROM orders ✓
Correct answer: D. SELECT only reads data; the others modify or destroy it.
A purely cosmetic text typo appears in the app's UI but nothing is broken. How is this typically classified?
  • A. Low severity, likely low priority ✓
  • B. High severity, high priority
  • C. Critical severity
  • D. P1 emergency
Correct answer: A. A cosmetic issue with no functional impact is normally low severity and low priority.
You've run through all your troubleshooting steps and still cannot resolve an issue. What is the right action?
  • A. Keep the customer waiting without updates
  • B. Escalate to L2/L3 with all details gathered so far ✓
  • C. Close the ticket as unresolved
  • D. Ask the customer to escalate it themselves
Correct answer: B. Escalating with full context lets the next tier continue without repeating your work.
A customer's app tries to connect to a PostgreSQL database on port 5432 but the connection times out. Which is a plausible cause?
  • A. The DNS name is misspelled
  • B. The browser cache is full
  • C. A firewall is blocking port 5432 ✓
  • D. The SSL certificate expired
Correct answer: C. A timeout on a specific port commonly means a firewall or security group is blocking that port.
A server log records an error at 13:30 UTC. A user in India (IST, UTC+5:30) says it happened at 7:00 PM. Do these refer to the same moment?
  • A. No, they are several hours apart
  • B. No, IST is behind UTC
  • C. It cannot be determined
  • D. Yes, 13:30 UTC equals 7:00 PM IST ✓
Correct answer: D. Adding 5 hours 30 minutes to 13:30 UTC gives 19:00 IST, which is 7:00 PM.
An API suddenly starts returning "429 Too Many Requests". What is happening?
  • A. The client is being rate-limited for sending too many requests ✓
  • B. The server has crashed
  • C. The endpoint has permanently moved
  • D. The auth token has expired
Correct answer: A. 429 signals the caller has exceeded the allowed request rate and is being throttled.
One service on a server is misbehaving while everything else runs fine. What is the least disruptive first action?
  • A. Reboot the entire server
  • B. Restart only the affected service ✓
  • C. Rebuild the server from scratch
  • D. Restore the whole system from backup
Correct answer: B. Restarting just the failing service minimizes impact on other services and users.
An outage is ongoing and you don't yet know the root cause. What's the best update to send the customer?
  • A. Send nothing until it is fully fixed
  • B. Send them the raw server stack trace
  • C. Acknowledge the issue, confirm you're investigating, and give a follow-up time ✓
  • D. Tell them another team is entirely at fault
Correct answer: C. Acknowledging and committing to a follow-up keeps the customer informed even before a fix exists.
One user keeps seeing an old version of a page while everyone else sees the updated one. What's a quick thing to try first?
  • A. Restart the web server
  • B. Change the DNS records
  • C. Escalate to engineering
  • D. Ask them to clear cache or do a hard refresh ✓
Correct answer: D. A single stale view is usually a local cache issue fixed by clearing cache or hard-refreshing.
You want to find where along the network path packets are being lost between a user and a server. Which tool helps?
  • A. traceroute or tracert ✓
  • B. nslookup
  • C. chmod
  • D. top
Correct answer: A. traceroute/tracert shows each hop on the path, revealing where packets are delayed or dropped.
A server can suddenly no longer write logs and several apps start failing. Which check should you run first?
  • A. Check the CPU temperature
  • B. Check disk space with df -h ✓
  • C. Check the monitor display settings
  • D. Check the keyboard layout
Correct answer: B. Inability to write files is a classic symptom of a full disk, so checking free space is the logical first step.

Medium round 30 questions

A customer reports your web app is 'down,' but your monitoring shows all servers healthy and other users are unaffected. What is the most effective first troubleshooting step?
  • A. Immediately restart the production application servers
  • B. Ask the customer for specifics: exact error, URL, browser, and when it started, ideally with a screenshot ✓
  • C. Tell the customer it's working fine on your end and close the ticket
  • D. Escalate directly to the senior backend engineering team
Correct answer: B. When symptoms aren't reproducible system-wide, gathering concrete reproduction details from the reporter is the fastest path to isolating whether it's environmental, user-specific, or a real bug.
You need to check whether a web server is returning the correct HTTP status and headers for a given URL from the command line. Which command is most appropriate?
  • A. ping example.com/api/health
  • B. curl -I https://example.com/api/health ✓
  • C. traceroute example.com
  • D. netstat -an | grep 443
Correct answer: B. curl -I sends a HEAD request and prints the response status line and headers, which is exactly what you need to inspect an endpoint's HTTP response.
A user gets a '403 Forbidden' error on a page that other users can access normally. Which cause is most consistent with this symptom?
  • A. The server is completely down and unreachable
  • B. The user's account lacks the permission or role required for that resource ✓
  • C. There is a DNS resolution failure for the domain
  • D. The database has run out of disk space
Correct answer: B. A 403 means the server understood the request but refuses to authorize it, which for a single affected user typically points to missing permissions rather than an outage or infrastructure failure.
While tailing application logs to diagnose an intermittent error, which command lets you watch new log lines appear in real time and filter for the word 'ERROR'?
  • A. cat app.log | grep ERROR
  • B. grep ERROR app.log | wc -l
  • C. tail -f app.log | grep ERROR ✓
  • D. head -n 100 app.log | grep ERROR
Correct answer: C. tail -f streams new lines as they are written, and piping to grep ERROR filters that live stream to only matching lines.
A customer says their API requests started failing with '401 Unauthorized' this morning, though nothing changed on their side. What is the most likely explanation to check first?
  • A. The API endpoint URL was permanently deleted
  • B. Their API key or access token expired or was rotated/revoked ✓
  • C. The customer's monitor has a rendering glitch
  • D. Their local machine ran out of RAM
Correct answer: B. A 401 indicates missing or invalid authentication credentials, and an expired or rotated token is the most common reason previously working requests suddenly fail with 401.
You suspect a slow page load is caused by one specific backend call. Which browser tool gives you the clearest per-request timing breakdown?
  • A. The browser's Console tab
  • B. The browser's Network tab (DevTools) ✓
  • C. The browser's bookmarks manager
  • D. The operating system's task manager
Correct answer: B. The Network tab in DevTools lists each request with its status, size, and timing waterfall, letting you pinpoint which call is slow.
A customer reports data they entered 'disappeared,' but you can see the record exists in the database with correct values. What is the most likely front-end explanation to investigate?
  • A. The database dropped the table and recreated it
  • B. The user is viewing a stale cached page or is logged into a different account/environment ✓
  • C. The server's CPU is overheating
  • D. The record was encrypted and cannot be shown
Correct answer: B. When the backend clearly has the data but the user can't see it, caching, session/account mismatch, or wrong environment are the usual causes to check first.
When writing a public reply to a frustrated customer whose issue you've confirmed is a real bug now scheduled for a fix, which response is best?
  • A. 'This isn't really a bug, you're just using it wrong.'
  • B. Acknowledge the impact, confirm it's a known bug, give a realistic ETA or workaround, and offer to follow up ✓
  • C. Promise it will be fixed within the hour to calm them down
  • D. Copy and paste a generic canned apology with no specifics
Correct answer: B. Good support acknowledges the customer's experience, sets honest expectations, and provides a concrete path forward, which builds trust without overpromising.
A SQL query a customer relies on is returning zero rows unexpectedly. You want to verify their data exists without risking any changes. Which query is safest to run first?
  • A. DELETE FROM orders WHERE customer_id = 42
  • B. UPDATE orders SET status = 'active' WHERE customer_id = 42
  • C. SELECT COUNT(*) FROM orders WHERE customer_id = 42 ✓
  • D. TRUNCATE TABLE orders
Correct answer: C. A SELECT COUNT(*) is read-only and safely confirms whether matching rows exist, whereas DELETE, UPDATE, and TRUNCATE all modify or destroy data.
A customer's requests are intermittently failing and you see '429 Too Many Requests' in their logs. What does this indicate?
  • A. The server crashed and needs a reboot
  • B. They are hitting a rate limit and should throttle or batch their requests ✓
  • C. Their SSL certificate is invalid
  • D. The request body was malformed
Correct answer: B. HTTP 429 signals the client has sent too many requests in a given window, so the fix is to respect the rate limit by slowing down or spacing out requests.
A single user reports a public website is very slow, but all other users and your monitoring show normal performance. What is the most sensible first step?
  • A. Restart the production web server
  • B. Investigate the user's local network, browser, and device ✓
  • C. Redeploy the application
  • D. Immediately escalate to the development team
Correct answer: B. When only one user is affected while metrics are normal, the issue most likely lies on the client side, so start there.
An HTTP 502 Bad Gateway error from a reverse proxy typically means what?
  • A. DNS resolution failed
  • B. The upstream server returned an invalid response to the proxy ✓
  • C. The client's request timed out
  • D. The SSL certificate expired
Correct answer: B. 502 means a gateway or proxy received an invalid or unexpected response from the upstream server.
Which command quickly identifies the process listening on TCP port 8080?
  • A. ping 8080
  • B. lsof -i :8080 ✓
  • C. cat /etc/hosts
  • D. df -h
Correct answer: B. lsof -i :8080 lists the process bound to that port; ss -ltnp works similarly.
A customer's API request returns HTTP 401. What does this indicate?
  • A. The rate limit was exceeded
  • B. Authentication failed or credentials are missing/invalid ✓
  • C. The endpoint does not exist
  • D. The server crashed
Correct answer: B. 401 Unauthorized means valid authentication credentials are missing or incorrect.
Which command finds all lines containing the word ERROR in a large application log file?
  • A. grep ERROR app.log ✓
  • B. chmod ERROR app.log
  • C. cat ERROR app.log
  • D. find ERROR app.log
Correct answer: A. grep searches a file for lines matching a pattern, here ERROR.
What does the command 'tail -f application.log' do?
  • A. Shows the last lines and streams new lines as they are appended ✓
  • B. Deletes the last lines of the file
  • C. Sorts the file alphabetically
  • D. Counts the number of lines
Correct answer: A. tail -f follows the file, printing new content in real time as it is written.
A server's application is failing due to a full disk. Which command best helps locate what is consuming space in the current directory?
  • A. du -sh * ✓
  • B. ps aux
  • C. top
  • D. whoami
Correct answer: A. du -sh * summarizes the size of each item in the current directory to find large consumers.
An API repeatedly returns HTTP 429. What is happening?
  • A. Access is forbidden
  • B. The client is being rate limited for too many requests ✓
  • C. The gateway timed out
  • D. The credentials are invalid
Correct answer: B. 429 Too Many Requests means the client has exceeded the allowed request rate.
A user can reach a site by typing its IP address but not by its domain name. What is the most likely cause?
  • A. An SSL certificate misconfiguration
  • B. A DNS resolution problem ✓
  • C. A firewall blocking port 443
  • D. The database being down
Correct answer: B. Working by IP but not by name points to a DNS resolution failure for that domain.
What is the key difference between HTTP 403 and HTTP 404?
  • A. 403 is a server error; 404 is a client error
  • B. 403 means access is forbidden; 404 means the resource was not found ✓
  • C. They are functionally identical
  • D. 403 is a redirect; 404 requires authentication
Correct answer: B. 403 Forbidden denies access to an existing resource, while 404 means the resource does not exist.
A user can successfully ping a website's IP address but cannot open the site by its domain name. What is the most likely cause?
  • A. The web server is down
  • B. A DNS resolution problem ✓
  • C. An expired SSL certificate
  • D. A firewall blocking ICMP
Correct answer: B. Reaching the IP but not the name points to name resolution (DNS) failing on the client's side.
A reverse proxy returns HTTP 502 Bad Gateway to users. What does this most directly indicate?
  • A. The client sent a malformed request
  • B. The proxy received an invalid response from the upstream backend ✓
  • C. The requested URL does not exist
  • D. The user is rate limited
Correct answer: B. 502 means the gateway/proxy got an invalid or no response from the upstream server it forwarded to.
You need to find which process is listening on port 8080 on a Linux host. Which command is appropriate?
  • A. ss -tlnp | grep 8080 ✓
  • B. df -h /8080
  • C. chmod 8080 file
  • D. ping localhost:8080
Correct answer: A. ss -tlnp lists listening TCP sockets with the owning process, filterable by port.
An application log repeatedly shows 'Connection refused' when trying to reach a database on port 5432. What is the most probable cause?
  • A. The database credentials are wrong
  • B. No process is listening on that port / the service is down ✓
  • C. The disk is full
  • D. The SSL certificate expired
Correct answer: B. 'Connection refused' means the target port had no listener, typically because the service is stopped or bound elsewhere.
Which grep invocation searches a log file for the word 'error' regardless of letter case?
  • A. grep -v error app.log
  • B. grep -c error app.log
  • C. grep -i error app.log ✓
  • D. grep -l error app.log
Correct answer: C. The -i flag makes the pattern match case-insensitively (Error, ERROR, error).
A customer suddenly starts receiving HTTP 429 responses from an API. What is happening?
  • A. Their authentication token expired
  • B. They are being rate limited for too many requests ✓
  • C. The endpoint was moved permanently
  • D. The server is under maintenance
Correct answer: B. 429 Too Many Requests indicates the client exceeded the API's allowed request rate.
Only one customer reports a web app is slow while all others are fine and monitoring is green. What is the best FIRST diagnostic step?
  • A. Restart the production web server
  • B. Reproduce from a different network/location and run traceroute from the user ✓
  • C. Roll back the last deployment
  • D. Increase server RAM
Correct answer: B. When the issue is isolated to one user, you first confirm whether the problem is local to their network/path before touching production.
A browser shows NET::ERR_CERT_DATE_INVALID when loading a site over HTTPS. What is the most likely root cause?
  • A. The DNS record was deleted
  • B. The TLS/SSL certificate has expired or is not yet valid ✓
  • C. The server ran out of disk space
  • D. The user's password is wrong
Correct answer: B. This error means the certificate's validity dates don't cover the current time, usually an expired cert.
Which SQL query returns the total number of rows in a table named orders?
  • A. SELECT COUNT(*) FROM orders; ✓
  • B. SELECT * FROM orders;
  • C. SELECT SUM(orders);
  • D. COUNT ROWS orders;
Correct answer: A. COUNT(*) returns the number of rows in the result set of the query.
When should a Support Engineer typically escalate a ticket to L2/engineering rather than continue troubleshooting?
  • A. As soon as any customer complains
  • B. When the issue requires code changes or access beyond the support tier's scope ✓
  • C. Whenever the ticket is more than one hour old
  • D. Only after the customer asks to escalate
Correct answer: B. Escalation is warranted when resolution needs privileges, code changes, or expertise outside the current tier's remit.

Hard round 30 questions

A service intermittently returns HTTP 502 from an Nginx reverse proxy sitting in front of an upstream Node.js app. `ss -s` on the proxy shows a growing number of sockets in TIME_WAIT, and `dmesg` is clean. The 502s correlate with traffic spikes. Which cause is most consistent with this evidence?
  • A. The upstream is closing keep-alive connections faster than Nginx expects, so Nginx sends requests on a half-closed connection and gets RST/EOF before a response ✓
  • B. The TLS certificate on the upstream has expired, causing handshake failures under load
  • C. The Linux conntrack table is full, silently dropping new connections to the upstream
  • D. DNS resolution for the upstream hostname is timing out during spikes
Correct answer: A. 502 Bad Gateway with EOF/RST on reused keep-alive sockets is the classic upstream-closed-idle-connection race, aggravated at high request rates.
You run `dig +trace example.com A` and it resolves correctly, but `dig example.com A @10.0.0.53` (your internal resolver) returns SERVFAIL. `dig example.com A @10.0.0.53 +cd` (checking disabled) succeeds. What is the most likely root cause?
  • A. The authoritative nameserver is rate-limiting the internal resolver
  • B. DNSSEC validation is failing on the internal resolver, likely due to a broken chain of trust or a clock skew problem ✓
  • C. The internal resolver has an empty root hints file
  • D. The A record TTL expired and the zone was not re-signed
Correct answer: B. SERVFAIL that disappears with +cd (checking disabled) is the signature of a DNSSEC validation failure on the validating resolver.
A Java service under load throws no OutOfMemoryError but latency P99 climbs from 40ms to 3s. The GC log shows frequent lines like `[Full GC (Ergonomics) ... 7800M->7750M(8000M), 5.2 secs]` recurring every few seconds. What does this pattern indicate?
  • A. Normal young-generation churn that can be ignored
  • B. A native memory leak outside the Java heap
  • C. The heap is nearly exhausted and the collector is in a GC-thrashing / 'GC overhead' state, reclaiming almost nothing per Full GC ✓
  • D. Metaspace is undersized and should be increased
Correct answer: C. Repeated multi-second Full GCs that free only ~50M of an 8G heap mean the live set nearly fills the heap, so the JVM thrashes reclaiming almost nothing.
A pod is repeatedly restarted. `kubectl describe pod` shows `Last State: Terminated, Reason: OOMKilled, Exit Code: 137`, and the container's `resources.limits.memory` is 512Mi. The app is a JVM with no `-Xmx` set. What is the most precise explanation?
  • A. Kubernetes killed the pod because a liveness probe failed
  • B. The JVM defaulted its max heap to a fraction of the 512Mi limit but total process memory (heap + metaspace + threads + native) exceeded 512Mi, so the kernel cgroup OOM-killed it ✓
  • C. The node ran out of memory and evicted the pod
  • D. Exit code 137 means the app called System.exit(137) on a fatal error
Correct answer: B. 137 = 128+9 (SIGKILL); the cgroup memory limit was breached because JVM total footprint (not just heap) exceeded 512Mi even with container-aware heap sizing.
Given this Postgres plan fragment for a slow query: `Seq Scan on orders (cost=0.00..48210.00 rows=1 width=8) Filter: (customer_id = 4821) Rows Removed by Filter: 2399999`. An index on `orders(customer_id)` already exists. Why might the planner still choose a sequential scan here?
  • A. Sequential scans are always faster than index scans in Postgres
  • B. Stale statistics cause the planner to mis-estimate selectivity, or the index is unusable (e.g., customer_id column type differs from the literal, disabling the index) ✓
  • C. The index is on the wrong table and cannot be used
  • D. Postgres never uses an index when the filter removes more than a million rows
Correct answer: B. The estimate `rows=1` vs 2.4M removed signals a bad estimate (stale ANALYZE) or a type mismatch preventing index use; correct estimates would favor the index for a single-row result.
During a tcpdump on a client that reports 'connection hangs', you see: client sends SYN, server replies SYN-ACK, client sends ACK, client sends a data segment, then the client retransmits that same data segment repeatedly with no ACK from the server, and finally an RST. The handshake completed fine. What is the most likely fault?
  • A. A firewall or middlebox is dropping established-connection data packets but allowed the handshake, or there is asymmetric routing so the server's ACKs never reach the client ✓
  • B. The server's DNS record points to the wrong IP
  • C. TLS negotiation failed because of a cipher mismatch
  • D. The client's TCP window size is set to zero
Correct answer: A. A completed handshake followed by unacknowledged data retransmits points to a stateful device/asymmetric path dropping data-carrying packets or return ACKs, not a name or TLS issue.
A curl to an HTTPS endpoint fails with `SSL certificate problem: unable to get local issuer certificate`, but the same URL loads fine in a browser on the same machine. What is the most accurate explanation?
  • A. The server certificate has expired
  • B. The server is not sending the intermediate CA certificate(s) in the chain; the browser fills the gap via cached/AIA-fetched intermediates, but curl's trust store only has the root and cannot build the chain ✓
  • C. The hostname does not match the certificate's SAN
  • D. curl does not support TLS 1.3 while the browser does
Correct answer: B. Browsers cache or fetch missing intermediates via AIA, masking a server that omits the intermediate cert; curl builds strictly from the presented chain plus roots, so it fails to find the issuer.
An L7 load balancer uses round-robin across 4 backends. One backend has a subtle bug that makes ~25% of requests return HTTP 500. Health checks (GET /healthz returning 200) pass. Users complain of sporadic errors. Which change best mitigates while you fix the root cause?
  • A. Switch the algorithm to least-connections
  • B. Enable sticky sessions so affected users stay on one backend
  • C. Make health checks deeper/outlier-detection based (eject a backend that returns a high rate of 5xx on real traffic), not just a static /healthz probe ✓
  • D. Increase the health check interval
Correct answer: C. A shallow /healthz masks a partially-failing backend; passive health checks / outlier detection on real 5xx rates will eject the bad instance from rotation.
A Kafka consumer group's lag is steadily growing on one partition while other partitions of the same topic stay near zero. Throughput of producers is uniform across keys. What is the most likely cause?
  • A. The broker holding that partition's leader is down
  • B. A single 'hot key' (or skewed partitioning) routes disproportionate messages to that partition, and/or one consumer instance handling it is slow/blocked on a poison message ✓
  • C. The consumer group's offset was reset to earliest
  • D. Replication factor is too low for that partition
Correct answer: B. Lag isolated to one partition while producer load is uniform points to key skew concentrating messages there or a single slow/stuck consumer for that partition.
You need to find, in a 20GB access log, the top 5 client IPs by request count only for requests that returned 503, where the IP is field 1 and status is field 9 (space-delimited). Which pipeline is correct and memory-efficient?
  • A. awk '$9==503{print $1}' access.log | sort | uniq -c | sort -rn | head -5 ✓
  • B. grep 503 access.log | awk '{print $1}' | head -5
  • C. awk '{print $1, $9}' access.log | grep 503 | wc -l
  • D. sort access.log | awk '$9=503{print $1}' | uniq | head -5
Correct answer: A. Only option A filters on the exact status field, counts per IP via sort|uniq -c, and orders descending; grep 503 would also match bytes/latency fields and `$9=503` is an assignment, not a comparison.
Users intermittently see SSL certificate mismatch errors on a site served by several servers behind a load balancer. What is the most likely cause?
  • A. The CA root is missing on every client
  • B. One node behind the load balancer serves a mismatched or expired certificate ✓
  • C. TLS 1.3 is not supported
  • D. HTTP/2 is disabled globally
Correct answer: B. Intermittent cert errors across a pool usually mean one back-end node has an outdated or wrong certificate.
A TCP connection remains in the TIME_WAIT state. What does this indicate?
  • A. The connection failed to establish
  • B. The socket that performed the active close is waiting to handle any delayed packets ✓
  • C. There is packet loss on the link
  • D. A firewall dropped the connection
Correct answer: B. TIME_WAIT is held by the side that actively closed the connection to safely absorb any late-arriving segments.
A curl request fails with 'Could not resolve host'. At which layer is the failure occurring?
  • A. TLS handshake
  • B. DNS resolution ✓
  • C. TCP connection refused
  • D. HTTP 500 response
Correct answer: B. 'Could not resolve host' occurs before any connection is made, indicating a DNS lookup failure.
An application returns HTTP 504 Gateway Timeout only under heavy load. What is the most likely root cause?
  • A. The client's browser cache is stale
  • B. The upstream server does not respond within the proxy's timeout window ✓
  • C. The SSL certificate expired
  • D. The DNS record is incorrect
Correct answer: B. 504 means the proxy/gateway gave up waiting for a slow upstream, common when backends are overloaded.
A Linux process appears in the 'Z' (zombie) state. What does that mean?
  • A. It is consuming 100% CPU
  • B. It has terminated but its exit status has not yet been read by its parent ✓
  • C. It is blocked on disk I/O
  • D. It is stuck in an infinite loop
Correct answer: B. A zombie has finished executing but lingers in the process table until its parent reaps it with wait().
Customers report that emails sent by your application are landing in spam. Which records most directly govern email authentication and deliverability?
  • A. A and CNAME records
  • B. SPF, DKIM, and DMARC records ✓
  • C. The MX record TTL
  • D. The PTR record only
Correct answer: B. SPF, DKIM, and DMARC authenticate the sending domain and strongly influence spam classification.
A client sees 'Connection refused' when connecting to a service. What does this specifically indicate?
  • A. A firewall silently dropped the packet
  • B. The host is reachable but nothing is listening on that port (an RST was returned) ✓
  • C. DNS resolution failed
  • D. The TLS negotiation failed
Correct answer: B. 'Connection refused' means the host is up and actively rejected the connection with a RST because no service listens on that port; a dropped packet would instead time out.
A database query that ran fine for months now times out after the table grew large. What should a support engineer investigate first?
  • A. The user's browser font settings
  • B. The query execution plan and whether an appropriate index is missing ✓
  • C. Restarting the client machine
  • D. Clearing the local DNS cache
Correct answer: B. Slowdowns that appear as data grows often stem from full table scans; checking the plan and indexes is the first move.
After a certificate expires on an HSTS-enabled site, users cannot click through the browser warning to reach it. Why?
  • A. The browser software is broken
  • B. HSTS instructs the browser to use only HTTPS and forbids bypassing certificate errors ✓
  • C. The HTTP port is closed
  • D. The DNS record is poisoned
Correct answer: B. HSTS enforces HTTPS-only access and removes the option to proceed past certificate errors.
A 4-core Linux server shows a 1-minute load average of 8.0. What does this most accurately mean?
  • A. The CPU is at 8% utilization
  • B. The system is idle
  • C. On average 8 tasks are competing for CPU while only 4 cores exist, indicating saturation ✓
  • D. RAM usage is 8 GB
Correct answer: C. Load average counts runnable/running tasks; 8.0 on 4 cores means roughly twice the demand the CPU can serve.
What is the correct sequence of the TCP three-way handshake?
  • A. ACK, SYN, SYN-ACK
  • B. SYN, SYN-ACK, ACK ✓
  • C. SYN-ACK, ACK, FIN
  • D. SYN, ACK, RST
Correct answer: B. The client sends SYN, the server replies SYN-ACK, and the client completes it with ACK.
What is the key semantic difference between HTTP 301 and HTTP 302 responses?
  • A. 301 is a client error, 302 is a server error
  • B. 301 is a permanent redirect, 302 is a temporary redirect ✓
  • C. 301 is cacheable, 302 can never be cached
  • D. 301 requires HTTPS, 302 does not
Correct answer: B. 301 signals the resource moved permanently (search engines update), while 302 indicates a temporary move.
In a traceroute, one intermediate hop shows high latency but the final destination hop shows normal low latency. What is the most accurate interpretation?
  • A. The destination server is overloaded
  • B. That intermediate router is likely deprioritizing ICMP replies and is not necessarily a real problem ✓
  • C. Packet loss is occurring at the destination
  • D. DNS resolution is failing at that hop
Correct answer: B. Routers often rate-limit or deprioritize ICMP, so a single high-latency middle hop with a healthy end-to-end result is usually cosmetic.
A 4-core Linux server shows a 1-minute load average of 4.00. What does this most accurately indicate?
  • A. The CPU is idle
  • B. The system is roughly fully utilized, about one runnable task per core ✓
  • C. The server is severely overloaded at 4x capacity
  • D. Memory is exhausted
Correct answer: B. Load average equal to the core count means, on average, each core has one runnable process, i.e., near full but not saturated.
Which set of HTTP methods is considered idempotent?
  • A. GET, PUT, DELETE ✓
  • B. POST, PATCH, GET
  • C. POST, PUT, DELETE
  • D. GET, POST, PATCH
Correct answer: A. GET, PUT, and DELETE produce the same server state when repeated; POST is not idempotent.
An application intermittently fails with 'Too many open files'. Which limit should you investigate first?
  • A. The disk quota (df)
  • B. The per-process file descriptor limit (ulimit -n) ✓
  • C. The maximum RAM (free -m)
  • D. The CPU core count (nproc)
Correct answer: B. This error is caused by hitting the file descriptor limit, controlled by ulimit -n / systemd LimitNOFILE.
Two database transactions hang and one is aborted with a 'deadlock detected' message. What condition causes this?
  • A. Both transactions read the same row simultaneously
  • B. Each transaction holds a lock the other needs, creating a cycle ✓
  • C. The database ran out of connections
  • D. A missing index slowed the query
Correct answer: B. A deadlock occurs when transactions form a circular wait, each holding a lock the other requires.
Before a planned IP migration, you lower a DNS record's TTL from 86400 to 300 a day in advance. What is the main effect of this change?
  • A. It permanently speeds up all future queries
  • B. Resolvers cache the record for a shorter time, so the later cutover propagates faster ✓
  • C. It encrypts the DNS response
  • D. It prevents the record from being cached at all
Correct answer: B. A lower TTL shortens caching so that when you change the record, clients pick up the new value quickly.
On Linux, a process was terminated unexpectedly and dmesg shows an 'Out of memory: Killed process' entry. What happened?
  • A. The process crashed due to a segmentation fault
  • B. The kernel OOM killer terminated it to reclaim memory under pressure ✓
  • C. The disk filled up
  • D. A user ran kill -9 manually
Correct answer: B. The kernel's OOM killer selects and terminates a process when the system runs critically low on memory.
What does the command 'curl -I https://example.com' do?
  • A. Downloads the full page body
  • B. Sends a HEAD request and returns only the response headers ✓
  • C. Ignores TLS certificate errors
  • D. Follows all redirects silently
Correct answer: B. curl -I issues an HTTP HEAD request, retrieving status line and headers without the response body.

Prep for another role

Questions are original, written and independently verified for HireHire's role interview quizzes. They reflect the kind of knowledge Support Engineer interviews test, not any specific company's questions. HireHire maps live tech & IT jobs across India, updated regularly. Last updated: August 2026.