99.9% SLA is marketing. Read the exclusions. Scheduled maintenance. Force majeure. "Issues outside our control." I have seen 99.9% offered with 30 minute monthly allowance and no financial remedy. Worse: credits capped at monthly fee prorated, which for a $5 VPS is pennies. Demand defined compensation or ignore the number. I enforce this in my contracts. Three sentences.
1. SLAs that matter have teeth
1.1) Actual compensation percentages
1.1.a) 10% per hour down
1.1.b) Full month refund at 99% or below
2. SLAs that don't
2.a) "Best effort" language
2.b) Credit only, no refund
2.c) Exclusions longer than inclusions
3. My personal checklist
3.I) Read the TOS
3.ii) Check if they define "availability"
3.iii) Verify if monitoring is independent
4. The thread topic
4.a) Agree with cleardmitri
4.b) 99.9% without compensation defined is worthless
Member
I run ZFS mirrors on auction servers from RackNerd. Their SLA is 99.5% with no compensation. I don't care. My data is on three machines in two DCs, synced nightly. SLA is about their infrastructure. Your redundancy is about your uptime. Specs: 4x 4TB HGST, $18/TB, RAID10 local, zfs send to offsite. Uptime measured in years, not their promise.
Member
I got 3 months free from a german host once 2day I still use them 4ever the sla said 99.9 but they went down 4 6 hours in march 2024 I opened ticket cited clause 4.2 they paid out in credit within 48 hours no argument the trick is finding hosts that actually honor it most just hope you dont read or dont ask
Hey folks, good discussion so far. Just a reminder to keep it focused on SLAs and personal experiences, not naming specific staff or doxxing support agents. @dancore thanks for sharing that positive outcome—if anyone wants to name the provider, please take it to DMs per our hosting recommendation policy. Thread stays open, but let's not turn this into a provider review list. Appreciate the technical depth so far.
99.5% with no compensation is basically "we tried"
Member
Exactly. And I chose that. RackNerd's $30/year plans don't pretend to be enterprise. I measure my own uptime with UptimeRobot on all three nodes. Last year: two nodes had 100%, one had a 14-minute blip during a kernel update I triggered myself. Their SLA could be 50% and I'd still sleep fine.
This is the gap in the thread. Cleardmitri's OP is about contracts for business use. Steveipw is describing a hobby/infra setup where he owns the risk. Both valid, different worlds.
For business: I keep a spreadsheet. Column H is "SLA teeth score." 0 = no compensation. 1 = credit only. 2 = cash refund possible. 3 = automatic payout without ticket. I've found exactly two providers at the $50-100/mo tier with a 3.
Member
Name them? Or DM me. I am always looking. My german host was Hetzner actually they have 99.9% SLA on cloud servers with 5% credit per hour over the limit max 100% monthly bill. Not automatic but I got it in 48 hours. Not a 3 by your scale but solid 2.
Per kat_fold's note, provider recommendations to DMs. But Hetzner's SLA is public info, fair to discuss. Their cloud SLA is indeed 5%/hour, capped at 100%, credit not refund. I've used them for client projects in Riga. The claim process is ticket-based but straightforward.
What I want to add: even "defined compensation" is worthless if the measurement is self-reported. Always run independent monitoring. I use Pingdom + a second source. Discrepancies happen more than you'd think.
