whatsssaaa

hey

45 W for all three is impressive. Which boxes are they?

Three small boxes under the desk:

• node-a: KStack control plane and DNS
• node-b: web and mail
• node-c: Standby, plus backups to an external disk

Total power draw is about 45 W idle. Photos and the Standby dashboard to follow.

Whichever you pick, put the panel and monitoring on a different provider from the customer servers. Learned that one the hard way.

We run both. Hetzner for price per core, OVH when a customer wants a specific region. Support response times are similar in our experience.

Starting a small hosting business and torn between Hetzner and OVH for the first few dedicated servers.

The things we care about: network quality into the UK, how fast support answers at night, and whether IPv6 is any trouble. What's your experience?

We'll be applying kernel updates to node4 on 3 November, 02:00–03:00 UTC.

Sites on that node will be unavailable for up to 10 minutes while it reboots. Mail is queued and delivered once the node is back. No action is needed on your side.

When a customer upgrades from a 2 GB to a 4 GB plan halfway through the month, what does Zephyr do with the difference: credit, immediate invoice or next renewal?

We'd prefer an immediate invoice for the difference.

The "no runbook, no page" rule is the one we're stealing.

Sharing the escalation policy that finally stopped us waking the whole team:

• Tier 1: the on-call engineer, paged straight away.
• Tier 2: a secondary after 10 minutes without an acknowledgement.
• Tier 3: the team lead after 25 minutes.

Alerts only page if they have a runbook attached. Everything else goes to a channel for the morning.

The panel lets us pick a PHP version per site, but the API only exposes it read-only. We'd like to bump a batch of sites from 8.2 to 8.3 from a script.

Anyone else need this?

Day to day NodeBB sits around 300 MB, so 1 GB looks generous. The asset build is a different story though. Webpack and the minifier spike well past that.

What do people run with? We've settled on 2 GB for any site that rebuilds themes or plugins.

Great write-up. Point 3 catches almost everyone.

We finished moving 40 WordPress sites off an old cPanel box last month. Notes for anyone planning the same:

1. Lower DNS TTLs a day ahead. 300 seconds made the cut-over painless.
2. Move mail separately. Sites first, mailboxes the following weekend.
3. Check wp-config.php for hard-coded paths. Three sites had /home/user/public_html baked in.
4. Use staging copies for the awkward ones. Anything with WooCommerce got a staging run first.

Happy to answer questions.

Has anyone got wildcard certificates working when the zone is on Cloudflare rather than our own nameservers?

The HTTP challenge is fine for single names, but *.example.com needs DNS-01. What token permissions does the panel need?

They follow the site's backup policy, so the retention you set there applies. They're labelled as push snapshots in the backup list, so they're easy to tell apart from the nightly ones.

Every Push back to production takes a snapshot of the live site first. How long do those stick around, and do they count against the backup quota?

We push a couple of times a week and I'd like to know how far back we can roll.

We used it to try this forum's new theme before it reached the real site. Next on the list: a diff view before pushing.

Also give the container at least 2 GB while NodeBB rebuilds. At 1 GB the asset build can run out of memory part-way, and visitors keep the old cached stylesheet.