Twelve RackNerd locations still answer on a public Looking Glass, and each of those pages prints the test IPv4 for its own node. RackNerd moved its DC-02 equipment into DC-03 in May 2026, and the DC-02 link most test-IP lists still hand out no longer resolves, so plenty of the addresses in circulation lead to pages that are gone. Every URL and IP below was opened and checked on November 15, 2026.
The Twelve Nodes That Answer
Each page returned a 200 response and showed its own Server Location and Test IPv4 field. The ASN is the network the address is announced from, pulled from the registry rather than from RackNerd's copy.
Location | Looking Glass | Test IPv4 | ASN |
|---|---|---|---|
Los Angeles DC-03 | 107.174.51.158 | AS36352 | |
San Jose, CA | 192.210.207.88 | AS36352 | |
Seattle, WA | 192.3.253.2 | AS36352 | |
Chicago, IL | 198.23.228.15 | AS36352 | |
Dallas, TX | 198.23.249.100 | AS36352 | |
New York | 192.3.81.8 | AS36352 | |
Ashburn, VA | 192.3.254.158 | AS36352 | |
Atlanta, GA | 107.173.164.160 | AS36352 | |
Toronto, ON | 167.160.186.8 | AS36352 | |
Amsterdam, NL | 23.94.101.88 | AS36352 | |
Strasbourg, FR | 45.95.173.2 | AS29066 | |
Dublin, IE | 23.95.225.2 | AS36352 |
Why lg-lax02 Keeps Showing Up in Other Lists
RackNerd's status page logged two relocation batches, on May 21 and May 22, 2026, that moved DC-02 equipment to its DC-03 facility at 1200 W 7th St. in downtown Los Angeles. The Looking Glass hostnames went with the hardware. lg-lax01, lg-lax02 and lg-lax05 now return NXDOMAIN from public resolvers, and 204.13.154.3, the address nearly every DC-02 list quotes, is registered to Multacom's AS35916 rather than the ColoCrossing space the current nodes use.
We will be beginning the first batch of our Los Angeles DC-02 to Los Angeles DC-03 physical migration project today. (RackNerd status page, May 21, 2026)
So when a list offers you a DC-01, DC-02 or DC-05 address, you are looking at a snapshot from before the move. DC-03 is the only Los Angeles entry left on RackNerd's own Looking Glass page, and the VPS order form labels it the same way.
Running a Test That Tells You Something
A single ping answers one question: can I reach this box right now. Four steps get you closer to what the server will feel like in production.
- Measure from your side. ping 107.174.51.158 first, then mtr -rwzc 20 107.174.51.158 on macOS or Linux. On Windows, pathping does the same job with more patience.
- Test the reverse direction. Open the node's Looking Glass, paste your own address into the destination field, pick MTR, and run it. The page prints Your IP Address near the top, which saves you a lookup when you are behind CGNAT.
- Measure throughput. curl -o /dev/null -s -w '%{speed_download}\n' https://lg-lax03.racknerd.com/test_files/100MB.test returns bytes per second. Every live node serves 10MB, 100MB and 1GB files under /test_files/.
- Repeat at the hour you care about. A 3 p.m. test on a quiet route says very little about 9 p.m. on the same route.
The Eight Datacenters Without a Looking Glass
RackNerd's datacenter page lists 20 locations, but the KVM VPS page names 12 that are instantly provisioned, and those 12 are exactly the nodes in the table above. The other eight have no Looking Glass page that I could find.
Location | Test IPv4 in third-party lists | Announced by | Status |
|---|---|---|---|
Utah | 104.156.142.1 | WebNX (AS18450) | Not on RackNerd's site |
New Jersey | 192.3.165.30 | ColoCrossing (AS36352) | Not on RackNerd's site |
Tampa / Miami | 209.216.90.1 | KUDUCOM (AS22549) | Same IP listed for both |
Montreal | 38.153.100.129 | Cogent (AS174) | Not on RackNerd's site |
London | 185.239.172.14 | Cogent (AS174) | Not on RackNerd's site |
Frankfurt | None published | — | No data |
Singapore | None published | — | No data |
Treat every address in that second table as unconfirmed. None of them appears on a RackNerd page I could reach, and Frankfurt and Singapore have nothing published at all. If you need one of those locations, ask support for a test address before you pay.
Which Node Should You Test First
- Serving users in Asia: start with Los Angeles DC-03, San Jose and Seattle. They are the three closest RackNerd nodes to the Pacific, and none of them is a China-optimized route.
- Serving users in Europe: Amsterdam, Strasbourg and Dublin. Strasbourg is the odd one out, since its test IP is announced from AS29066 while the other eleven nodes sit in AS36352.
- Serving users on the US East Coast: Ashburn, New York and Atlanta, with Chicago and Dallas as the middle of the country.
- Reading the labels: the New York node's address geolocates to Buffalo in public databases. RackNerd's label names the region, not a street address, so do not expect Manhattan latency.
Once a node passes, the cheapest way in is the 1 GB KVM special at $21.99/year, with 2 GB at $35.99/year and 4 GB at $59.99/year (checked on RackNerd's specials page, November 15, 2026). The order form prints the test IP beside each location in the dropdown, and for the 1 GB plan that dropdown currently offers New York, Los Angeles DC-03, Dallas and Seattle, so stock varies by plan.
Questions Before You Order
Do I need an account to use the RackNerd Looking Glass?
No. The page is open, the destination field takes a hostname or an IP, and Ping, Traceroute and MTR run without a login.
Why will lg-lax02.racknerd.com not load?
That hostname was tied to the DC-02 facility, and RackNerd moved that equipment to DC-03 in May 2026. The current Los Angeles page is lg-lax03.racknerd.com, and its test IP is 107.174.51.158.
Can I run IPv6 tests against these nodes?
No IPv6 test addresses are published on the Looking Glass pages. RackNerd hands out IPv6 blocks on request in Los Angeles and France, up to 100 addresses per VPS, but you have to open a ticket after ordering.
Is the test IP the address my VPS will get?
No. It is a host in the same facility on the same upstream, which is enough to judge routing and latency. Your VPS is assigned a different address from RackNerd's pool.
Which RackNerd location has the lowest latency for me?
Only your own traceroute can answer that. Test two or three nodes from the table at the hour your users are active, compare packet loss as well as average latency, and pick the one that holds up under load.
0 comments
No comments yet
Leave a comment