Internet connection SLA checker (RTT & Packet Loss)

A small Bash tool to verify an access SLA by measuring ICMP Round-Trip Time (RTT) and Packet Loss (PL) towards a set of neutral reference targets (e.g., RIPE Atlas anchors hosted at Internet Exchange Points).

It is intended for:

• Customer-side checks (internal validation / troubleshooting)
• Evidence for tickets opened to an ISP when the guaranteed SLA is not met

The script saves raw measurement logs plus a machine-readable summary into a bundle directory that can be attached to a support request.

What changed in the current version

Compared to the original version, the script is now more suitable for SLA verification:

• No root required by default (root is only useful to use shorter ping intervals)
• Packet loss is computed from TX/RX counters (not from the rounded “X% packet loss” field)
• RTT keeps decimals (no integer truncation)
• Per-target SLA evaluation (if any neutral target violates the threshold, the SLA is marked as KO)
• Generates a result bundle directory with:
  • raw per-target ping logs
  • meta.txt (environment + run parameters)
  • summary.json (easy to ingest or paste into a ticket)
• Optional progress bar while ping runs (useful because ping -q is silent until completion)

Requirements

• Bash
• ping
• awk, grep, wc, tr
• python3 is not required to run the script (it was only used to apply edits during development)

The script works on common Linux distributions and macOS. On systems without ping6, it tries ping -6.

Usage

./SLA.sh <icmp_packets> <rtt_threshold_ms> <loss_threshold_pct>

Example (acceptance test: RTT <= 50 ms, loss <= 0.02%):

./SLA.sh 1000 50 0.02

Notes:

• With -q (quiet), ping prints only a summary at the end. The script therefore shows a progress bar per target (estimated from packets * interval).
• For a quick sanity check use small counts (e.g., 20–50). For a more solid sample, use 1000 or more. Very large counts may take a long time.

Output bundle (attach to ISP tickets)

Each run creates a directory (default name includes timestamp), for example:

sla_results_2026-01-13T20-58-56+0100/
  meta.txt
  summary.json
  ping_v4_193.201.40.210.log
  ping_v4_217.29.76.27.log
  ping_v6_2001_7f8_10_f00c__210.log
  ...

• meta.txt includes: OS, ping version (if available), thresholds, packet count, interval, target list, run ID.
• ping_*.log are raw outputs from ping (useful as evidence).
• summary.json contains a compact result summary (status + averages for reporting).

SLA decision logic

The SLA is evaluated per target:

• KO if for any target:
  • RTT average is missing (often indicates 100% loss / no replies), or
  • RTT average is greater than the threshold, or
  • packet loss is greater than the threshold
• OK only if all configured targets meet the thresholds

Averages reported in summary.json are for convenience; they do not override the per-target decision.

Configuration

Targets are defined at the top of SLA.sh:

AHv4=( "193.201.40.210" "217.29.76.27" )
AHv6=( "2001:7f8:10:f00c::210" "2001:1ac0:0:200:0:a5d1:6004:27" )

You can edit/add targets as needed.

Environment variables

You can override some runtime parameters without editing the script:

• SLA_OUTDIR — set the output directory name/path
• SLA_INTERVAL — seconds between probes (default: 0.2, root default: 0.11)
• SLA_SIZE — ICMP payload size (default: 56)
• SLA_PROGRESS — set to 0 to disable the progress bar

Example:

SLA_OUTDIR=/tmp/sla_run_001 SLA_INTERVAL=0.5 SLA_PROGRESS=0 ./SLA.sh 1000 50 0.02

Notes and limitations

• Some networks and remote hosts may deprioritize or filter ICMP. If ICMP is filtered, results may be inconclusive even if user traffic works.
• The tool measures RTT to the chosen neutral targets, not end-to-end application performance.
• If IPv6 is not available, the IPv6 section is reported as INCONCLUSIVE (unless your SLA explicitly requires IPv6, in which case treat it as KO at the process/policy level).

License

See LICENSE.

Credits

Original project: SLA check script by SBTAP-AS59715, with thanks to @mphilosopher.