Q: The system is experiencing unexplained network latency. `ping` times internally jump from 1ms to 200ms periodically. How do you isolate the problem?
Intermittent high latency is tricky. My workflow would be:
Want to master this scenario in a live sandbox? The Linux Foundation's Certified System Administrator (LFCS) & SRE Track covers this exact problem with hands-on terminal drills.
🛠️ Production Runbook & Step-by-Step Resolution
Direct Answer: "Never reboot a server blindly; always capture top process telemetry, lsof descriptors, and thread dumps first. The interviewer is testing: Advanced network troubleshooting, packet capture, kernel networking stack.. I structure my answer around systematic triage first, root cause analysis second, and permanent remediation third."
Initial Diagnostics & Root Cause Analysis
Intermittent high latency is tricky. My workflow would be:
- Scope: Is it affecting all instances? One AZ? One specific app? I'd ping the default gateway and another instance in the same subnet to isolate whether it's the host network stack or the physical network/hypervisor.
- Host metrics: Check
dmesgorjournalctlfor NIC errors, ring buffer drops, ornf_conntracktable full messages. - Traffic Analysis: Use
mtrto track packet loss at hops. To capture the issue, I'd runtcpdump -i eth0 -w capture.pcapduring the latency spikes.
Remediation & Permanent Safeguards
Based on the finding, I might tune ring buffers (ethtool -G), adjust sysctl buffers, or escalate to the cloud provider if the host metrics check out perfectly.
- Kernel Drops: Use
netstat -suto see if UDP/TCP packets are being dropped by the kernel due to full receive buffers. I'd verifytopto check if a specific CPU core handling NIC interrupts (ksoftirqd) is pinned at 100%, causing a processing bottleneck.
- Scope: Is it affecting all instances? One AZ? One specific app? I'd ping the default gateway and ...
- Host metrics: Check dmesg or journalctl for NIC errors, ring buffer drops, or nf_conntrack table ...
- Traffic Analysis: Use mtr to track packet loss at hops. To capture the issue, I'd run tcpdump -i ...
Senior hiring managers and bar raisers evaluate how you prioritize during outages. Avoid these common anti-patterns:
- Issuing immediate kill -9 (SIGKILL) commands rather than allowing graceful termination via kill -15 (SIGTERM) when configuring system experiencing unexplained SIGKILL immediately halts execution without allowing processes to flush memory buffers, commit database transactions, or remove stale PID locks.
-
Overlooking file system inode exhaustion (df -i) when diagnosing 'No space left on device' errors when configuring system experiencing unexplained
Millions of small files or unrotated session tokens can exhaust available file system inodes even when standard disk storage (
df -h) shows ample free space. - Exhausting open file descriptor allocations (ulimit -n) under intensive concurrent socket connections when configuring system experiencing unexplained Reaching file descriptor exhaustion throws 'Too many open files' errors, preventing services from accepting new TCP client socket connections.
-
Permitting defunct zombie processes to accumulate without fixing parent process SIGCHLD waitpid reaping when configuring system experiencing unexplained
Zombie processes consume process table entries; accumulating zombies can exhaust the system maximum PID limit (
pid_max).