Campaign and e-commerce peaks, load balancer in place
The highest-fit case after the fully prepared one: the load balancer exists, the peaks are known, and the only missing feature is planning for them. Scheduled scaling is a Pro feature; everything else, including self-healing and the cost guard, is in every plan. A campaign peak that runs past the node capacity of your plan is created anyway — the capacity is priced, not enforced against your traffic.
- Prerequisite work
- none; wants scheduled scaling
- Time to set up
- 20 min
- Adapter
- Snapshot or webhook
- Fit
- 9/10
Prerequisites, time and fit are this case's row in the onboarding cases table. Hetscale's own setup is about 15 minutes in every case; the rest is work on your side.
Steps
- 01Connect read-only
The day-0 report from your last 30 days shows the peaks, how often capacity fell short and a ranged cost. Campaign days with CPU pinned at 100% are flagged.
- 02Rehearse, then go Live
One Rehearsal, then a range wide enough for the next campaign: minimum 2, maximum what the report suggests for the peak.
- 03Schedule the known peaks
Scheduled scaling (Pro and above) raises the minimum before a launch or a sale so the nodes are already healthy when the traffic arrives, instead of reacting to it.
- 04Know what happens when the type sells out
Campaign days are when a server type is most likely to be out of stock in your location. Hetscale then tries the other locations of your network zone; if you allow it, larger types fill in until your type is back: up to three you list in your order, or the cheapest one in stock within a price cap you set.
FAQ for this case
What happens if the server type is out of stock during the sale?
Another location of your network zone is used. If none has the type, a group in a fallback mode creates one of the types you allowed, the first in stock from your list or the cheapest in stock within your price cap, at the difference its settings show, and replaces it with your type once it is back; otherwise the group is marked in capacity deficit and you are notified. Stock is observed centrally: a create is skipped while the reading says out of stock, and tried now and then in case the reading is wrong.
Will scaling in cut a customer’s checkout?
Scale-in removes the node from the load balancer first and waits the drain window (default 60 s, up to 300 s) before shutting it down. Raise the window if checkouts hold long connections.
What if the sale needs more nodes than my plan covers?
They are created. The node capacity of a plan is a commercial number, never the reason a campaign runs short: the scale-out runs, bounded by your maximum, the cost guard, your euro ceiling, your own node cap if you set one, and the next plan’s capacity. Time spent above the capacity counts against a monthly budget; when it is spent, one mail gives you seven days to move up a plan. Nothing running is ever touched, and the plan never holds healing.