MicrosaaS owners need better disaster recovery options
16 posts across 7 communities in the last 60 days; 31% from r/selfhosted.
“Specific: Has anyone else noticed a decline in Sol capacity the past couple days? Namely, it doesn't seem to be taking time to "think" anymore, and by my watch it was always pretty dumb if it didn't take at least a few seconds. Also, I used to be able to prompt-correct it by…”
“I wanted to PM you but i cant, your SLA are a little too high i believe, 99.5, 99 would be more realistically, considering they will give you 1 to 5 downtimes. maybe i can give u a hand?”
“Just reference - http://uptime.is/ Also don't just ChatGPT entire posts based on 1 thought that everyone already knows [re: 99.9% uptime sounds great until you do the math: it allows 8h 46m of downtime a year, and your real number is probably lower]”
Several users express concern about lacking disaster recovery (DR) plans and seek alternatives. They question existing solutions and discuss the difficulty of implementing robust DR strategies, particularly for smaller operations.
SOURCES (19)
“Just reference - http://uptime.is/ Also don't just ChatGPT entire posts based on 1 thought that everyone already knows [re: 99.9% uptime sounds great until you do the math: it allows 8h 46m of downtime a year, and your real number is probably lower]”
“I'd treat a few complaints as expected from that thin cohort. Kill on the gate anyway, leave a one-click restore for two weeks, and only reverse if people outside the original 10% actually restore it in volume.”
“Specific: Has anyone else noticed a decline in Sol capacity the past couple days? Namely, it doesn't seem to be taking time to "think" anymore, and by my watch it was always pretty dumb if it didn't take at least a few seconds. Also, I used to be able to prompt-correct it by saying "think about this for a few" if it gave a quick/dumb answer, and that doesn't appear to work now. (also getting a lot of "Could not load this ChatGPT conversation" in the inte”
“Botched an update of a multimaster DB that held actual money, lot of it. Can not back such things up in normal way, only hope of recovery are transaction logs. Didn't have to go that far, but it was a dicey half an hour.”
“are you planning to solve the single-SMB-pool dependency before rebuilding the services, or is that staying as the foundation?”
“Adding that his backup software should be able to backup locally then add Wasabi as the immutable cloud tier.”
“I wanted to PM you but i cant, your SLA are a little too high i believe, 99.5, 99 would be more realistically, considering they will give you 1 to 5 downtimes. maybe i can give u a hand?”
“I made a 20 year career out of being 1 day ahead of the client in my tiny self hosted environment. If I met my teen self tomorrow, I would tell them IT is an artistic hobby and certainly not a career. Times have changed so quickly. With any luck the price will encourage overproduction of all these parts and within 2 years we're in the opposite economy flooded with yesteryears components at discount prices. Those datacenters are either upgrading or going out of business. I'll scavenge eit”
Wow, struck a nerve. Calm down friend, it's not that deep.
“yet when AWS and cloudflare go down, it results in millions of damage and loss..”
“The internet as we know it is changing for the worse. Yes -- and that's actually one of the things why I am reconsidering, but I see it a bit different than you. I fear that cyberattacks will become much worse and if you don't follow up patching and inspecting systems 24/7 I would become a victim. I don't even trust that being behind two routers and only reachable via VPN is absolute protection. My approach is now shifting a bit towards: - Cloud for non-privacy relevant stuff or if i”
“There are definitely private hosted models? It's much simpler for most orgs to just pay for something like https://tinfoil.sh , or setup a ZDR contract, instead of hosting their own multi-terabyte model at scale. And yea I know this is locallama, and they should host when it makes sense, but it often doesn't.”
“Honestly, their storage/bandwidth bill must be astronomical.. They probably host an incredible amount of garbage models.. I wonder if their business is profitable? But I think I'll download few models just in case..”
“Have backups, restore them on a VPS. Connect VPS home to my router via VPN (until you have new hardware).”
“Imho Hardware failure is best to be handled by a secondary server equipped with recent copy of data). this server is always off and just running to pull new copy. Also not reachable from outside home network.”
“Then a 45 minute run leaves you 15 minutes of margin, and one slow restore turns hourly into two-hourly”
“Absolutely, having a single source of truth is the only way to get everyone on board.”
“This is a bad idea during a hardware crisis. And regardless of the crisis, people will absolutely blame and expect you to support them when things go wrong. This is why PC builders have a hard time.”