Thank you for everyone's replies! I think we're going to > @chimichurri said:
I'm afraid Azure too is above your stated budget...
400GB of bandwidth alone is 35 dollars, add to that a crappy B1S (1 vCPU burstable, 10% baseline) instance...
(and you said you need 100 GB of storage too..)
For all Virtual Machines that have two or more instances deployed across two or more Availability Zones in the same Azure region, we guarantee you will have Virtual Machine Connectivity to at least one instance at least 99.99% of the time.
(otherwise it's just 99.9% at most)
For any Single Instance Virtual Machine using Premium SSD or Ultra Disk for all Operating System Disks and Data Disks, we guarantee you will have Virtual Machine Connectivity of at least 99.9%.
Yeah, you're right. I was looking at the wrong section for transfer.
Actually if it's just one server you're looking for (and not let say the same x10), and you'd be OK with either a 1/8 of an Epyc or up to couple of ARM cores / VPS, perhaps just go with Oracle Cloud?
That's a top contender right now. They've got the E4's for $20/m or so and the free tier covers everything else. 99.99% durability, live migrations.
Alternative (albeit similar) idea, provided Europe would also be tolerable: Hetzner! (has a lot of street cred in these circles and arguably provides what you're looking for)
Doesn't work for the same reason as Clouvider : (
Thank you for everyone's responses! We'll take a closer look at OVH's public cloud, Oracle, and Upcloud. If everyone has any experience with those, I'd love to hear.
@chimichurri said: Possible downside: Bandwidth might be limited only up to 50Mbps (based on various YABS samples for EPYC instances, example ; unsure about ARM)
OVH used to have CEPH backed VPS (auto migration in case something happens to your host node) but I don't know if they have that by now. Maybe on the "public cloud" line. Hetzner has it in EU but for now not in US. Maybe that will arrive sometime.
What are you trying to do anyway? Maybe you want to pursue some more serious high availability scheme. Multiple VPS with floating IP's, anycast routing to multiple DC's, etc. Cloudflare Workers might be another viable approach.
@willie said:
OVH used to have CEPH backed VPS (auto migration in case something happens to your host node) but I don't know if they have that by now. Maybe on the "public cloud" line. Hetzner has it in EU but for now not in US. Maybe that will arrive sometime.
OVH public cloud uses Ceph, yeah - we actually used them for ~ 3 months, and moved away because IO latency performance was too lackluster actually, coupled with a few network stability issues now. We're on Upcloud right now, which is been decent so far.
What are you trying to do anyway? Maybe you want to pursue some more serious high availability scheme. Multiple VPS with floating IP's, anycast routing to multiple DC's, etc. Cloudflare Workers might be another viable approach.
That is obviously the ideal solution - they're internal-facing monitoring nodes - but Zabbix isn't something that easily runs in HA (done it before). At least, not until 6.0 is stable. Time/maintenance cost there doesn't make a lot of sense compared to a instance that can hold a 99.99%+ yearly standard, as we have backup systems for the remaining 52 minutes
I have generally found dedicated servers to have less random downtime than VPS of any sort. VPS operators are always rebooting the host node for something, etc. So for your purposes, maybe try a small Kimsufi.
@willie said:
I have generally found dedicated servers to have less random downtime than VPS of any sort. VPS operators are always rebooting the host node for something, etc. So for your purposes, maybe try a small Kimsufi.
For lower-range providers, probably - but most mid-high range providers - eg, OVH, 3 Big Clouds, etc will have live migrations for host-node maintenance. In theory, virtualized should offer better reliability. Honestly, though, that's sometimes not the case.
@johnk said: will have live migrations for host-node maintenance
Exactly. A UK provider I've been with for a while I know does this. They did a write up about it when mitigating the intel bugs (spectre etc) and I've got a VPS with them that just hit 4 years uptime.
They were bought out by iomart so the technical side (blog posts, technical staff to speak too, etc) has all gone unfortunately but I did manage a DC tour before hand lol.
@willie said:
That was the idea of CEPH backed VPS (…)
Live migration does not require Ceph. Live migration works with KVM and any form of local storage as well. It‘s a bit slow, because it has to copy the entire VMs disk, but for the purpose of upgrading host nodes where you can schedule the migration upfront that’s not really an issue.
The advantage of Ceph is replication over multiple servers instead of just multiple disks (as with RAID). That‘s a core component for High-Availability, as you can immediately restart a VM on another hypervisor and don’t need to try to manually recover the VMs data from the crashed node. But Ceph isn’t a requirement for live migrations and zero-downtime host upgrades.
We are indeed phasing out Ceph-based cloud products. If anyone has any detailed questions about them, please write to our support team directly. The technicians on the cloud team will be able to better help you.
We're Katie and Lea and we'll do our best to answer questions you have about Hetzner Online. We and not our employer are responsible for any horrible puns and dated cultural references.
Comments
Hetzner has some ceph based vms, but thats EU.
Didnt clouvider expand to the USA recently-ish? Might be worth to ask them directly. @Clouvider
Hey teamacc. You're a dick. (c) Jon Biloh, 2020.
They've got dedis but the OnApp vps line is still UK only.
Thank you for everyone's replies! I think we're going to > @chimichurri said:
Yeah, you're right. I was looking at the wrong section for transfer.
That's a top contender right now. They've got the E4's for $20/m or so and the free tier covers everything else. 99.99% durability, live migrations.
Doesn't work for the same reason as Clouvider : (
Thank you for everyone's responses! We'll take a closer look at OVH's public cloud, Oracle, and Upcloud. If everyone has any experience with those, I'd love to hear.
1Gbps public bandwidth on 4core ARM instance.
Would give Oracle a miss...seemed like amateur hour last time I tried them
Interesting...thanks for sharing - if you don't mind elaborating, what issues did you hit with them?
try innoscale.net
OVH used to have CEPH backed VPS (auto migration in case something happens to your host node) but I don't know if they have that by now. Maybe on the "public cloud" line. Hetzner has it in EU but for now not in US. Maybe that will arrive sometime.
What are you trying to do anyway? Maybe you want to pursue some more serious high availability scheme. Multiple VPS with floating IP's, anycast routing to multiple DC's, etc. Cloudflare Workers might be another viable approach.
OVH public cloud uses Ceph, yeah - we actually used them for ~ 3 months, and moved away because IO latency performance was too lackluster actually, coupled with a few network stability issues now. We're on Upcloud right now, which is been decent so far.
That is obviously the ideal solution - they're internal-facing monitoring nodes - but Zabbix isn't something that easily runs in HA (done it before). At least, not until 6.0 is stable. Time/maintenance cost there doesn't make a lot of sense compared to a instance that can hold a 99.99%+ yearly standard, as we have backup systems for the remaining 52 minutes
It looks like hetzner are depreciating ceph on the cloud
Get the best deal on your next VPS or Shared/Reseller hosting from RacknerdTracker.com - The original aff garden.
https://docs.hetzner.com/cloud/servers/faq/#what-is-the-difference-between-local-and-ceph-disks-for-servers-on-the-cx-models
Contribute your idling VPS/dedi (link), Android (link) or iOS (link) devices to medical research
Hetzner or Lunanode?
I have generally found dedicated servers to have less random downtime than VPS of any sort. VPS operators are always rebooting the host node for something, etc. So for your purposes, maybe try a small Kimsufi.
What about Ramnode?
I was with them from 2013-2019 when Nick was running it and they were superb - especially support.
But now since they got bought, I'm not sure if they're still as good as they were before.
For lower-range providers, probably - but most mid-high range providers - eg, OVH, 3 Big Clouds, etc will have live migrations for host-node maintenance. In theory, virtualized should offer better reliability. Honestly, though, that's sometimes not the case.
Exactly. A UK provider I've been with for a while I know does this. They did a write up about it when mitigating the intel bugs (spectre etc) and I've got a VPS with them that just hit 4 years uptime.
They were bought out by iomart so the technical side (blog posts, technical staff to speak too, etc) has all gone unfortunately but I did manage a DC tour before hand lol.
That was the idea of CEPH backed VPS but the performance hit is significant. It was interesting to hear that Hetzner has deprecated it now.
Live migration does not require Ceph. Live migration works with KVM and any form of local storage as well. It‘s a bit slow, because it has to copy the entire VMs disk, but for the purpose of upgrading host nodes where you can schedule the migration upfront that’s not really an issue.
The advantage of Ceph is replication over multiple servers instead of just multiple disks (as with RAID). That‘s a core component for High-Availability, as you can immediately restart a VM on another hypervisor and don’t need to try to manually recover the VMs data from the crashed node. But Ceph isn’t a requirement for live migrations and zero-downtime host upgrades.
— Michael
Alwyzon - Virtual Servers in Austria starting at 4,49 €/month (excl. VAT)
https://novos.be/ is using ceph
Amadex • Hosting Forums • Wie ist meine IP-Adresse? • AS215325
Forum for System Administrators: sysadminforum.com
We are indeed phasing out Ceph-based cloud products. If anyone has any detailed questions about them, please write to our support team directly. The technicians on the cloud team will be able to better help you.
We're Katie and Lea and we'll do our best to answer questions you have about Hetzner Online. We and not our employer are responsible for any horrible puns and dated cultural references.