Conversation

Notices

1. Embed this notice
   Phantasm (phnt@fluffytail.org)'s status on Saturday, 18-Jan-2025 00:58:07 JST Phantasm

   • 7666
   @7666 :francisco: franned.it
   • ✙ dcc :pedomustdie: :phear_slackware: likes this.
   • kuteboiCoder repeated this.
   • Embed this notice
     Phantasm (phnt@fluffytail.org)'s status on Saturday, 18-Jan-2025 01:00:20 JST Phantasm

     in reply to

     • 7666
     @7666 I'm honestly having my doubts about BuyVMs future without him, because he said multiple times he isn't as interested in running a VPs provider as he used to. His main thing now is Namecrane (email and cpanel hosting)
     kuteboiCoder likes this.
   • Embed this notice
     kuteboiCoder (kuteboicoder@subs4social.xyz)'s status on Saturday, 18-Jan-2025 01:00:40 JST kuteboiCoder

     in reply to

     • 7666

     @phnt@fluffytail.org @7666@comp.lain.la
     
     I feel like I should do a better job of keeping abreast of the cloud platforms available

   • Embed this notice
     7666 (7666@comp.lain.la)'s status on Saturday, 18-Jan-2025 01:57:58 JST 7666

     in reply to
     @phnt Namecrane is a registrar too, which is super interesting. Most likely he is chasing the money because running services on TOP of his existing infra extracts way more revenue. Being a registrar, offering managed services, email hosting, reseller services, etc. All much higher margin than just straight compute, but arguably requires more elbow grease to get right, so with capital inflow from dumping BuyVM he's blowing right past the capital requirements to do all that now.
     
     I still hate that he threw in the towel though. "Cloudzy AI". What a fucking name.
     ✙ dcc :pedomustdie: :phear_slackware: likes this.
   • Embed this notice
     7666 (7666@comp.lain.la)'s status on Saturday, 18-Jan-2025 01:58:51 JST 7666

     in reply to
     @phnt well my current services suck ass outside of the fat pipes so i'm just exiting anyway. for YEARS i have tried to have robust caching at the edge and get continually frustrated by shitty performance. I literally got francisco to make me a custom plan with more SSD space so I could lvmcache the garbage slabs.
     
     no more. i blow $10k and francisco is gone, and i already decided to blow it. fuck it
     ✙ dcc :pedomustdie: :phear_slackware: likes this.
   • Embed this notice
     7666 (7666@comp.lain.la)'s status on Saturday, 18-Jan-2025 01:58:52 JST 7666

     in reply to
     @phnt I'm not even worried about data residency. Even though I cut checks for Francisco to the tune of $4,000 a year, I still expect this guy to show up and start poking his head around and go "This guy who is pushing like 600TB a month needs to pay up". His other business has strict bandwidth caps and getting more bandwidth is like, prohibitively expensive.
   • Embed this notice
     Phantasm (phnt@fluffytail.org)'s status on Saturday, 18-Jan-2025 01:58:52 JST Phantasm

     in reply to

     • 7666
     @7666 I wouldn't worry about paying up for bandwidth _for your current services_. The contract they signed includes a clause where BuyVM has to have the same policies as they did before acquisition with the same pricing. However new services might be priced differently.
     
     It might be that the same pricing and features must also apply to newly bought services, but I would have to read the hundreds of messages on Discord again for that.
     obrazek.png
   • Embed this notice
     Phantasm (phnt@fluffytail.org)'s status on Saturday, 18-Jan-2025 01:58:53 JST Phantasm

     in reply to

     • 7666
     @7666 The owner of Cloudzy AI (and .com) and Fran know each other since before BuyVM existed and BuyVM is using some of their infra already, so it isn't a completely random company that bought the brand name. However since Cloudzy is operated from UAE a lot of question around UAE laws and US laws come into question. What about free speech, government involvement in infiltrating the infra and much more. And since neither Fran nor Hannah (the owner of Cloudzy) answered any question about that for more than a week, I'm kinda worried. UAE isn't really the perfect country where you want to have your hosting business registered.
   • Embed this notice
     Phantasm (phnt@fluffytail.org)'s status on Saturday, 18-Jan-2025 02:00:23 JST Phantasm

     in reply to

     • 7666
     @7666 I'm not moving completely away yet. If the LUX nodes end up in NL, I'm moving those immediately, the rest will stay unless Hannan proves he's a nontransparent money seeking retard like I sort of expect.
     
     If I end up migrating completely, it will be a massive pain in the ass and self-hosting at home isn't really a painless option either thanks to the EU being stupid like they've always been.
     ✙ dcc :pedomustdie: :phear_slackware: likes this.
   • Embed this notice
     Phantasm (phnt@fluffytail.org)'s status on Saturday, 18-Jan-2025 04:24:04 JST Phantasm

     in reply to

     • 7666
     @7666 Also about the Miami storage slab being completely dead and corrupted. From what Fran said in Discord, they only have a single slab node there, so if the FS (probably ZFS) can't recover at least some of it, it's probably gone. The way he said it, implies to me that there's basically no backup other than the RAID 10 redundancy.
     ✙ dcc :pedomustdie: :phear_slackware: likes this.
   • Embed this notice
     7666 (7666@comp.lain.la)'s status on Saturday, 18-Jan-2025 04:24:08 JST 7666

     in reply to
     @phnt Here's what it looks like for me. sda is the slab, vda is the local disk. ALL I/O just drops occasionally.
     
     I am so done with :francisco_intensifies:
   • Embed this notice
     Phantasm (phnt@fluffytail.org)'s status on Saturday, 18-Jan-2025 04:24:09 JST Phantasm

     in reply to

     • 7666
     @7666 It's probably an IOPS quota over some time frame, many VPS providers do that. OVH being one of the worst offenders. After around 2GB, their additional disk throttle to 12,10,8,4 MB/s of sustained read or write. Slowly throttling down as you keep using the disk.
     
     What I usually see is an almost increase in disk backlog to around 5-10 seconds for 15-20 minutes and then it goes back to normal. A reliable way to trigger it is to open a few tabs of random bigger threads, or make many search queries. It's not related to an increase in network traffic/requests, dmcache deciding to write the dirty pages to the primary disk or PostgreSQL deciding to do DB maintenance. It just happens out of nowhere unprovoked.
     number of IOPS by type and disk
     Backlog by disk
   • Embed this notice
     Phantasm (phnt@fluffytail.org)'s status on Saturday, 18-Jan-2025 04:24:10 JST Phantasm

     in reply to

     • 7666
     @7666 I have a 20GB DB and a 12 GB lvmcache in front of it. Basically everything hits the cache, but it chokes anyway (both the slab and the OS SSD).
   • Embed this notice
     7666 (7666@comp.lain.la)'s status on Saturday, 18-Jan-2025 04:24:10 JST 7666

     in reply to
     @phnt i'm running on a 3TB slab with a 425GB lvmcache. And yeah, it chokes due to what I believe are arbitary IOPS or storage bandwidth limits.
     
     What I end up seeing is a complete drop-out of the storage for upwards of 15 seconds and then all the queued I/O flushes at once. rinse and repeat.
     
     Better yet, if you don't use the storage for like a whole day, the iowait is basically nil. Then out of the blue, the issue comes back!
   • Embed this notice
     7666 (7666@comp.lain.la)'s status on Saturday, 18-Jan-2025 04:24:11 JST 7666

     in reply to
     @phnt >DB on a slab
     
     ok well i think he specifically says NOT to do that. it was in a ticket somewhere that i can't find
   • Embed this notice
     Phantasm (phnt@fluffytail.org)'s status on Saturday, 18-Jan-2025 04:24:12 JST Phantasm

     in reply to

     • 7666
     @7666 Yeah, I've been fighting performance issues since I moved the Pleroma DB to a slab and there's basically no way to solve it. It just chokes randomly for a few minutes multiple times a day and comes back. Even the OS SSD sometimes just gives up.
   • Embed this notice
     7666 (7666@comp.lain.la)'s status on Saturday, 18-Jan-2025 04:24:13 JST 7666

     in reply to
     @phnt there's a whole multitude of reasons why i'm moving but consistent performance and less points of failure are the big ones.