Question 680
A Linux server begins swapping heavily and user response time degrades even though CPU is not fully used. Which bottleneck is likely?
- /proc/PID/smaps_rollup
- Memory allocation and OOM logs
- Memory leak
- Memory pressure causing disk-backed paging
Explanation
Correct Answer
D:Swapping moves memory pages between RAM and backing storage, adding latency when working sets exceed available memory. Identify competing processes and working-set demands.
Incorrect Answers
A:smaps_rollup provides summarized memory mappings and residency information for a process; the requirement is diagnosing slowdown from memory paging, outside the effect of this choice.
B:OOM messages show allocation context and selected processes; it does not deliver diagnosing slowdown from memory paging, which is decisive in this scenario.
C:A leak retains memory that is no longer needed, gradually increasing working-set pressure; the evidence calls for diagnosing slowdown from memory paging, not merely this separate operation.
Question 681
A performance regression starts after a service release but system metrics do not explain it. Which next step gives application-level insight?
- iostat -xz 1
- sysstat sar historical reports
- Distributed tracing or request timing instrumentation
- Correlate application timing with block-device latency metrics
- ss -tpn socket listing
Explanation
Correct Answer
C:Request-level traces and latency spans reveal slow code paths and dependencies. Host CPU and memory charts cannot identify every application bottleneck.
Incorrect Answers
A:Extended iostat reports device throughput, utilization and latency-related statistics by interval; that is not equivalent to examining latency inside an application request.
B:sar can retrieve time-series CPU, memory and I/O activity when collection is enabled; the administrator instead needs examining latency inside an application request; this option serves a different purpose.
D:Matching request latency with I/O latency, throughput and queue depth helps identify storage stalls; that changes a different part of the system; the requested outcome is examining latency inside an application request.
E:ss can associate TCP sockets with owning processes when permissions allow; this does not achieve examining latency inside an application request on the system described.
Question 682
free shows little unused memory, but large buffers/cache and substantial MemAvailable. What is the correct interpretation?
- Memory pressure causing disk-backed paging
- Page cache reclaim under memory pressure
- Reclaimable cache is not necessarily memory exhaustion
- Memory allocation and OOM logs
Explanation
Correct Answer
C:Linux uses spare RAM for caching and can reclaim much of it when applications demand space. MemAvailable is more useful than the raw free column alone.
Incorrect Answers
A:Swapping moves memory pages between RAM and backing storage, adding latency when working sets exceed available memory; it will not provide interpreting Linux memory cache accurately under these conditions.
B:Linux may reclaim page-cache pages to satisfy memory demands; using it would leave interpreting Linux memory cache accurately unresolved.
D:OOM messages show allocation context and selected processes; it does not deliver interpreting Linux memory cache accurately, which is decisive in this scenario.
Question 683
A rotating disk is at nearly 100% utilization while several workloads queue requests. What resource may be saturated?
- IOPS and access pattern limit
- Disk I/O service capacity
- Idle CPU time with outstanding I/O
- Open deleted file retaining disk blocks
Explanation
Correct Answer
B:A busy storage device can accumulate latency as outstanding operations wait for service. Queue depth and device metrics help confirm saturation.
Incorrect Answers
A:Random small operations stress per-request latency and IOPS rather than maximum sequential bandwidth; the task involves diagnosing storage queue saturation, whereas this addresses another subsystem.
C:iowait reflects CPU idle intervals when at least one I/O request remains outstanding under Linux accounting; that is not equivalent to diagnosing storage queue saturation.
D:Unlinked files remain allocated while open file descriptors reference their inodes; the administrator instead needs diagnosing storage queue saturation; this option serves a different purpose.
Question 684
An engineer must establish whether disk reads, not CPU execution, dominate a request's runtime. Which method is appropriate?
- Distributed tracing or request timing instrumentation
- perf record and perf report
- iostat -xz 1
- Correlate application timing with block-device latency metrics
Explanation
Correct Answer
D:Matching request latency with I/O latency, throughput and queue depth helps identify storage stalls. One metric alone can mislead about causality.
Incorrect Answers
A:Request-level traces and latency spans reveal slow code paths and dependencies; that changes a different part of the system; the requested outcome is connecting application stalls with device response time.
B:Linux perf can sample hardware and software performance events to identify costly call paths; this does not achieve connecting application stalls with device response time on the system described.
C:Extended iostat reports device throughput, utilization and latency-related statistics by interval; it will not provide connecting application stalls with device response time under these conditions.
Question 685
A disk benchmark reports high sequential bandwidth but the application waits on tiny synchronous writes. What should be measured next?
- Small synchronous write latency
- Disk I/O service capacity
- IOPS and access pattern limit
- Compare df filesystem fullness with du directory sizes
Explanation
Correct Answer
A:Sequential megabytes per second do not characterize durability-sensitive small writes. Measure per-operation latency under representative queue depth.
Incorrect Answers
B:A busy storage device can accumulate latency as outstanding operations wait for service; the evidence calls for assessing fsync-heavy small-write performance, not merely this separate operation.
C:Random small operations stress per-request latency and IOPS rather than maximum sequential bandwidth; the task involves assessing fsync-heavy small-write performance, whereas this addresses another subsystem.
D:df shows allocated filesystem capacity while du attributes visible path consumption; its scope differs from assessing fsync-heavy small-write performance, so it cannot meet the stated requirement.
Question 686
Several disks show uneven utilization and the team needs per-device throughput and latency in short intervals. Which command fits?
- Optimize the measured hot function after correctness review
- iostat -xz 1
- Correlate application timing with block-device latency metrics
- pidstat -rud -p PID 1
Explanation
Correct Answer
B:Extended iostat reports device throughput, utilization and latency-related statistics by interval. It helps identify a specific busy block device.
Incorrect Answers
A:Profiling identifies where CPU time is spent, supporting targeted improvement; the administrator instead needs profiling storage performance per block device; this option serves a different purpose.
C:Matching request latency with I/O latency, throughput and queue depth helps identify storage stalls; that changes a different part of the system; the requested outcome is profiling storage performance per block device.
D:pidstat can sample process CPU, memory and I/O activity by PID over time; this does not achieve profiling storage performance per block device on the system described.
Question 687
The system reports high iowait while CPU cores are mostly idle. What does this metric suggest?
- Idle CPU time with outstanding I/O
- High storage request latency
- Open deleted file retaining disk blocks
- Small synchronous write latency
- Network or NFS server I/O congestion
Explanation
Correct Answer
A:iowait reflects CPU idle intervals when at least one I/O request remains outstanding under Linux accounting. It is not a precise per-process disk utilization measurement.
Incorrect Answers
B:Elevated await reflects time I/O requests spend queued and serviced; the requirement is interpreting high CPU iowait, outside the effect of this choice.
C:Unlinked files remain allocated while open file descriptors reference their inodes; it does not deliver interpreting high CPU iowait, which is decisive in this scenario.
D:Sequential megabytes per second do not characterize durability-sensitive small writes; the evidence calls for interpreting high CPU iowait, not merely this separate operation.
E:Remote storage latency depends on both network transport and server disk service; the task involves interpreting high CPU iowait, whereas this addresses another subsystem.
Question 688
An administrator needs per-thread CPU statistics for a multi-threaded process, not just a single aggregate PID value. Which view helps?
- Load includes tasks waiting in uninterruptible sleep
- taskset
- Context-switch rate from vmstat or pidstat
- top with thread display enabled
Explanation
Correct Answer
D:Thread views reveal individual hot threads hidden by per-process totals. This can identify a single-thread bottleneck.
Incorrect Answers
A:Linux load average counts runnable tasks and certain uninterruptible waits; its scope differs from finding hot threads within a multi-threaded process, so it cannot meet the stated requirement.
B:taskset controls the CPU affinity mask of a process; that is not equivalent to finding hot threads within a multi-threaded process.
C:Very high context-switch rates can indicate scheduling overhead or contention; the administrator instead needs finding hot threads within a multi-threaded process; this option serves a different purpose.
Question 689
A process is suspected of holding many network connections; the administrator needs per-process sockets. Which evidence is useful?
- ss -tpn socket listing
- Correlate application timing with block-device latency metrics
- Optimize the measured hot function after correctness review
- iostat -xz 1
- perf record and perf report
- sysstat sar historical reports
Explanation
Correct Answer
A:ss can associate TCP sockets with owning processes when permissions allow. It helps tie network activity to specific executables.
Incorrect Answers
B:Matching request latency with I/O latency, throughput and queue depth helps identify storage stalls; it will not provide mapping live TCP connections to process owners under these conditions.
C:Profiling identifies where CPU time is spent, supporting targeted improvement; using it would leave mapping live TCP connections to process owners unresolved.
D:Extended iostat reports device throughput, utilization and latency-related statistics by interval; the requirement is mapping live TCP connections to process owners, outside the effect of this choice.
E:Linux perf can sample hardware and software performance events to identify costly call paths; it does not deliver mapping live TCP connections to process owners, which is decisive in this scenario.
F:sar can retrieve time-series CPU, memory and I/O activity when collection is enabled; the evidence calls for mapping live TCP connections to process owners, not merely this separate operation.
Question 690
An incident review requires ongoing per-process resource sampling, per-device disk latency measurements, and historical system activity. Which THREE tools match these separate needs? Choose THREE.
- Correlate application timing with block-device latency metrics
- pidstat -rud -p PID 1
- iostat -xz 1
- sysstat sar historical reports
- ss -tpn socket listing
- perf record and perf report
Explanation
Correct Answers
B:pidstat can sample process CPU, memory and I/O activity by PID over time. It provides more scope than one instantaneous process listing.
C:iostat -xz 1 samples extended block-device metrics, including throughput, utilization and request-latency indicators. It identifies the storage-device side of the incident alongside pidstat process samples and sar history.
D:With collection enabled, sar reports historical CPU, memory, disk and other system activity. It adds a retrospective timeline that a live pidstat process sample or current iostat device view cannot supply.
Incorrect Answers
A:Matching request latency with I/O latency, throughput and queue depth helps identify storage stalls; the evidence calls for capturing process metrics, block device latency and historic host trends, not merely this separate operation.
E:ss can associate TCP sockets with owning processes when permissions allow; the administrator instead needs capturing process metrics, block device latency and historic host trends; this option serves a different purpose.
F:Linux perf can sample hardware and software performance events to identify costly call paths; it answers a neighboring question rather than capturing process metrics, block device latency and historic host trends.