Updated 2026-09-11. Written by Christopher Gilliland, founder of ZynVoss LLC — Fort Pierce, FL. Not a guarantee of uptime and not an SLA unless we signed one.
Someone texts “the site is down.” That sentence covers five different failures. If you treat all five as a code bug, you will republish a working site onto a dead disk, or you will rewrite a homepage while the name still points at last year’s host.
This is the first hour we run on our own box and on a client’s. Times are Eastern. The shop is at 6515 Spanish Lakes Blvd. The public site is zynvoss.com. We host on a Windows box with IIS and a dedicated volume for the live files. Your stack may differ. The order still holds.
Minute 0–5: prove it from outside the building
Open the URL on a phone that is not on the shop Wi-Fi. Then try the same URL on a second network if you have one. A PC that also hosts the site will lie to you. Hosts files, hairpin NAT, and “it loads for me” waste the next forty minutes.
Write down four facts before you touch files:
- Homepage HTTP status (200, 301, 403, 502, 503, timeout).
- A second public page (about, contact, a product). If only one URL fails, it is not “the site.”
- Login or portal, if you have one. A 200 homepage and a 500 login is a different job than a dead origin.
- The clock. “Down since lunch” and “down since 9:02” are not the same incident.
If the phone loads the site and the office PC does not, stop. That is local DNS, a proxy, or a browser with a stale HSTS pin. It is not a deploy.
Minute 5–15: name, certificate, then the box
If the name does not resolve, or it resolves to the wrong IP, do not edit HTML. Registrar and DNS clicks are a person with the account. On our side the public address has been 73.124.156.192 — we still confirm with a public echo before we accuse DNS. A new hostname without an A record will HTTP on the office PC (hosts file) and fail in the real world. We have that scar on a product subdomain. It is not mysterious.
If HTTPS warns and HTTP works, the site is up and the certificate is wrong or missing. That is win-acme / Let’s Encrypt or the registrar’s SSL product. It is not a rewrite of the homepage.
Then the box:
- Is the web service running? For us: IIS sites Started, app pools Started.
- Is the disk that holds the site mounted and writable? For us that is the
Z:volume. A missing volume looks like “the internet is down.” - Free space. We keep a floor of 5 GB on the site volume. A full disk makes IIS do stupid things.
Minute 15–30: one site, or the whole machine
On a box that hosts more than one site, check the neighbors. If www is 200 and the portal is 503, recycle the portal pool. Do not bounce the public site to fix a login. If everything on the box is dead, look at the machine: sleep settings, a Windows update reboot, a full C: drive, a NIC that dropped. The public internet can be fine while the host is not.
Read the last deploy. A portal publish that copied over production secrets, a physical path that now points at a backup drive, or an app pool left Stopped after a failed publish — those are the boring causes. We keep a rollback folder for that reason.
Logs next, not Slack. IIS status codes, the app’s own log, the disk event. A 500 with a stack beat a theory.
Minute 30–60: say what it is, then fix that thing
Pick one sentence and send it to the person who called:
- Certificate. HTTP works. HTTPS will look broken in Chrome until the cert is reissued.
- App pool stopped. Public pages may still serve static files. Login will not.
- Wrong folder. The site is serving an old copy. Path goes back to the live root.
- DNS. The name is not here. Do not publish while the name still points somewhere else.
- Code. One URL 500s. We have a stack and a revert.
Do not tell a customer “we are investigating performance” when the pool is stopped. Do not take a product off the site because an unrelated hostname failed. Do not click your own ads to “test.”
What this is not
- A promise of five-nines. We do not sell that sentence without a contract.
- A reason to rebuild the homepage during an outage.
- Proof we are SOC 2 or “bank-grade.” We run the checks above. That is the control.
Florida notes
Storm week takes out power and the cable modem before it takes out HTML. If the host is in the office, a UPS buys you minutes, not a day. After-hours phone trees fail for the same reason — see after-hours answering. If the site is the only way a crew gets tomorrow’s jobs, print a paper fallback once, while the site is still up.
Headquarters: Fort Pierce, FL 34951. Office (772) 971-4448. Customer support (772) 971-4448 (voice agent, not a human desk). Email [email protected]. Independent Florida LLC. Not affiliated with the ZYN nicotine brand.