The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
Things that broke in production (and the checks they became)
Backups that backed up nothing. The script dumped an empty Postgres database while n8n was actually running on SQLite, and the restore test didn't catch it. Now there are online SQLite snapshots, counts read before every dump, a 35-check restore test and a corruption drill that must fail.
A webhook down for five minutes. Recreating a container dropped n8n's manual attachment to the stack network. The network is now declared in compose.
Mail going to spam. After a network change, DKIM signing silently stopped. Lesson: test a mail fix on the path real mail takes, not from the host.
Every incident becomes a check in the static suite, which now stands at 1,400+ automated checks.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
Your contact form works. The emails just never arrive.
This week a booking form on a client site sent its notifications with PHP mail(). The owner's Gmail sent them to spam, or dropped them.
Before you blame the form, check how the email is sent:
• SPF: a DNS record listing which servers may send for your domain
• DKIM: each email is signed, so the inbox can verify it came from you
• DMARC: tells inboxes what to do when SPF or DKIM fails
• Send from your own domain over SMTP or an email API, not the server's mail()
• Test with a real inbox, and check the spam folder too
Where do your site's form emails go: an inbox, a CRM, or both?
One real site, one real tip, every day. Follow for more videos.
More builds: youtube.com/@sellwrightstudio · instagram.com/sellwrightstudio
This one catches so many sites. Moving form notifications off PHP mail() to authenticated SMTP with proper SPF and DKIM fixes most of these cases.
10 checks before you launch
Launching a site this week? Run through these first:
Every form sends, and you receive the email
Favicon and social share image are set
Page titles and meta descriptions are written
The sitemap is submitted in Google Search Console
Analytics is recording visits
A 404 page exists and links back home
Old URLs redirect to their new pages
HTTPS everywhere, with no mixed-content warnings
Tested on a real phone, not just a narrow browser window
Backups are running
Save this for your next launch. Which one gets forgotten most? My vote: number 7.
#WebsiteLaunch #LaunchChecklist #WebDevelopment #Startup #WebsiteTips
Such a handy launch checklist.
A schedule that retries blindly will double-charge, double-email, and double-write.
Before you flip on a cron job (n8n, Railway cron, or a Next.js scheduled route), run this checklist:
1. Stable idempotency key — one key per logical job (order_id + day), not a random UUID
2. Unique constraint in Postgres — conflict means skip, not a boolean race
3. Claim, then work — mark claimed in the same transaction as the key insert
4. Backoff with a ceiling — jitter + max attempts, then park it
5. Timeout every external call — a hung cron is worse than a failed one
6. Overlap lock — if the previous run is still live, skip or queue
7. Replay-safe logs + dead-letter — alert on failure, replay still respects the same key
Read-only syncs can retry freely. Payments, emails, CRM writes, and Slack posts need a key first.
Which of these do your scheduled jobs still skip?
