TCP Ping — Reachability & Latency Test
TCP Ping opens a real TCP connection to a host on a given port (443 by default) and tells you whether it is reachable and how long the round trip takes. Unlike a classic ping it does not use ICMP echo — it measures the connection a service actually depends on, which many hosts allow even when they block ICMP. Enter a host, optionally set a port, and get a plain reachable / not-reachable verdict with latency.
Reviewed by the SysGlass team · Updated July 21, 2026
What it checks
- Opens a real TCP connection to the host on the chosen port (default 443) and reports whether the connection succeeds
- Measures round-trip latency of the connection and reports min / avg / max in milliseconds
- Runs several connection attempts and reports loss as a percentage (e.g. 1 of 4 failed = 25%)
- Returns a one-line verdict (PASS / WARN / FAIL) alongside the raw reached, loss, and latency rows
- Shows the equivalent command (nc -vz host port) so you can reproduce the same test from your own shell
- This is a TCP connect test, not an ICMP echo — true ICMP ping and traceroute run from the SysGlass remote Looking Glass nodes
How to read the result
- PASS (reachable): the host accepted the TCP connection on that port with 0% loss; the average latency is shown in milliseconds
- WARN: the host is reachable but at least one attempt failed (partial loss) — a sign of intermittent connectivity, congestion, or a flaky firewall path
- FAIL: 100% loss / unreachable — nothing is listening on that port, a firewall is blocking it, or the host is down
- The latency row shows min / avg / max across the attempts; lower is better, and a wide min-to-max gap indicates jitter even if the average looks fine
- The loss row shows successful vs total attempts, so you can tell a one-off drop from total failure
- A FAIL means only that host+port is closed from this vantage point — the same host may still answer on a different port or from another network
Frequently asked questions
Is this a real ping (ICMP)?
No. It is a TCP connect test: it opens a TCP connection to the chosen port instead of sending an ICMP echo request. Many hosts and firewalls drop ICMP but still accept TCP, so a TCP ping is often a more reliable signal of whether a specific service is reachable. For true ICMP ping and traceroute, SysGlass runs those from its remote Looking Glass nodes.
What port does it test by default?
Port 443 (HTTPS). You can set any port you like. Choose the port of the service you actually care about — for example 22 for SSH, 25/465/587 for SMTP, 80 for HTTP, 3306 for MySQL, or 6379 for Redis. The test only tells you about the port you specify.
Why does it say unreachable when the website clearly works?
The TCP ping only checks one port. If you test a port where nothing is listening, or one a firewall blocks, it reports FAIL even though the site is up on another port. Re-test using the correct port — usually 443 (HTTPS) or 80 (HTTP) for a website.
What latency counts as good?
There is no fixed threshold; it depends on distance and the network path. Connections within the same region are often in the single-digit to low-tens of milliseconds, while cross-continent paths can exceed 100 ms. Watch the min/avg/max spread too — a wide gap points to jitter even when the average looks acceptable.
What does the loss percentage mean?
The tool makes several connection attempts and reports the share that failed. 0% is healthy. Partial loss (a WARN) suggests intermittent reachability, congestion, or an inconsistent firewall. 100% loss means the port could not be reached on any attempt and is reported as a FAIL.
Do I need to install anything or sign up?
No. It is a free public web tool — enter a host and an optional port and run it from the browser. The connection test is performed server-side by SysGlass, and the page also shows the equivalent nc -vz command if you want to repeat it locally.