When working with NFS-mounted directories, developers often need granular visibility into the I/O patterns of specific mounts. Standard system monitoring tools like iostat or vmstat provide system-wide statistics but lack mount-point specificity. Here's how to achieve precise NFS client-side monitoring.

The most effective approach combines kernel statistics with custom parsing:

# Method 1: Using /proc/self/mountstats
cat /proc/self/mountstats | grep -A 3 "/your/mount/point"

# Sample output parsing:
nfs_io_monitor() {
  local mountpoint=$1
  while true; do
    awk -v mp="$mountpoint" '$0 ~ mp {getline; getline; print $3" "$4}' /proc/self/mountstats
    sleep 1
  done
}

For more sophisticated monitoring, consider this Python script:

import time

def monitor_nfs_io(mount_path, interval=1):
    prev_stats = {}
    while True:
        with open('/proc/self/mountstats') as f:
            found = False
            for line in f:
                if mount_path in line:
                    found = True
                    continue
                if found and 'bytes' in line:
                    current = line.strip().split()
                    if prev_stats:
                        read_diff = int(current[0]) - prev_stats['read']
                        write_diff = int(current[1]) - prev_stats['write']
                        print(f"Read: {read_diff/interval/1024/1024:.2f} MB/s")
                        print(f"Write: {write_diff/interval/1024/1024:.2f} MB/s")
                    prev_stats = {'read': int(current[0]), 'write': int(current[1])}
                    found = False
                    break
        time.sleep(interval)

For RHEL-based systems, the nfs-utils package provides nfsiostat:

# Install if needed
sudo yum install nfs-utils

# Usage for specific mount
nfsiostat /mnt/nfs/share 1

When implementing continuous monitoring:

• Disk I/O overhead from frequent stats collection
• Impact of monitoring interval on accuracy
• Potential need for kernel-based tracing (eBPF) in high-performance environments

For long-term monitoring, consider integrating with tools like:

# Telegraf configuration example
[[inputs.procstat]]
  pattern = "nfsiostat"
  interval = "10s"
  metric_prefix = "nfs_stats"

When working with NFS mounts, monitoring specific I/O activity is crucial for performance tuning and troubleshooting. Unlike local filesystems, NFS introduces network overhead, making targeted monitoring essential.

The nfsiostat tool (part of nfs-utils) provides NFS-specific statistics similar to iostat:

# Install if needed
sudo apt-get install nfs-common  # Debian/Ubuntu
sudo yum install nfs-utils       # RHEL/CentOS

# Basic usage
nfsiostat /mnt/nfs_mount 1 5

Sample output columns include:

• ops/s - Operations per second
• kB_read/s - Read throughput
• kB_wrtn/s - Write throughput

For more granular control, parse /proc/self/mountstats:

#!/bin/bash

NFS_MOUNT="/mnt/nfs_share"

get_nfs_io() {
    awk -v mount="$NFS_MOUNT" '
    $1 ~ /^READ:/ {read_bytes=$2; read_ops=$5}
    $1 ~ /^WRITE:/ {write_bytes=$2; write_ops=$5}
    END {
        printf "READ: %.2f MB/s (%d ops)\n", read_bytes/1048576, read_ops;
        printf "WRITE: %.2f MB/s (%d ops)\n", write_bytes/1048576, write_ops;
    }' <(grep -A4 "statvers" /proc/self/mountstats | grep -A10 "$mount")
}

while true; do
    get_nfs_io
    sleep 1
done

For production environments, consider the node_exporter NFS collector:

# Enable NFS collector in node_exporter
./node_exporter --collector.nfs

# Sample PromQL query
rate(node_nfs_bytes_read[1m]) / 1024 / 1024  # MB/s read
rate(node_nfs_bytes_written[1m]) / 1024 / 1024  # MB/s written

For deep inspection, use this SystemTap script (requires kernel headers):

probe nfs.read {
    printf("NFS READ: %d bytes\n", $size)
    @reads[execname()] <<< $size
}

probe nfs.write {
    printf("NFS WRITE: %d bytes\n", $size)
    @writes[execname()] <<< $size
}

probe timer.s(1) {
    println("\nREAD Statistics (MB/s):")
    foreach (proc in @reads) {
        printf("%s: %.2f\n", proc, @sum(@reads[proc])/1048576)
    }
    
    println("\nWRITE Statistics (MB/s):")
    foreach (proc in @writes) {
        printf("%s: %.2f\n", proc, @sum(@writes[proc])/1048576)
    }
    
    clear(@reads)
    clear(@writes)
}

• Monitor during peak and off-peak hours for baseline comparison
• Correlate NFS metrics with network statistics (iftop, nethogs)
• Watch for NFS version differences (v3 vs v4 performance characteristics vary)