For my critical stuff, the truthful answer is that I use Azure.
For other production-quality stuff, I've had particularly good reliability with Avoro, UltraVPS.eu and Inception (at least in Europe).
If you configure LunaNode block storage as the boot volume, that can be pretty good because they reattach to a new node in the event of a problem. There can be some tricks with this (like avoiding UUIDs in the network config) but quite reliable.
@cybertech said:
i would take a look at avoro (now owned by php-friends)
Is their uptime high? My priority is uptime. Cloudvider is one of the best uptime providers I tried, im looking for alternative offers.
i had one for 2 or 3 years, apart from planned migration (xeon gold to epyc) and planned network upgrade, there was one or 2 network downtime that was unplanned. machine uptime to my knowledge was 99.999% (really gut feel, not monitored)
I went back and looked at my own stats for providers with > 6 months monitoring history.
The following had 100% uptime in the past 6 months: Avoro (Frankfurt), UltraVPS.eu (Düsseldorf), Inception (Amsterdam), GreenCloud (Dallas, Chicago), BuyVM (Las Vegas), HostHatch (Amsterdam), vServer.site (Düsseldorf), VirMach (various)
It is probably worth noting that the interval is 5 mins, so short outages may not be captured (seems like BuyVM is a bit susceptible to that). However, some of these are also monitored with Hetrix at a 1 minute interval and the stats there were generally in very close agreement.
Providers can differ a lot by location. HostHatch in Los Angeles is under 99% (poor), Inception in Phoenix is lower than in Europe (maybe 99.95% range), and UltraVPS.eu in Los Angeles had an outage in the past week which took them under 99.9%.
Mission critical stuff is on GCP, Cloudflare (workers/KV) and bunnyCDN. Generally try to engineer things to not use a VM cause GCP is expensive for that. So more of a blend of CF workers, cloud functions etc - think serverless vibes.
Katapult.io and fly.io have me intrigued but haven't built anything on it yet.
I'd probably do vultr because they seem to do well in comparisons, or katapult if feeling adventurous.
The LES servers I've got (couple of Nexus ones) are mostly testing rather than prod. But they seem good on uptime too
@havoc said: Katapult.io and fly.io have me intrigued but haven't built anything on it yet.
I am disappointed by fly's failover algorithm. The servers seem to go down more often than they should, which results in a "black spot" until BGP updates and a new instance starts. Then I end up with two servers in a single region, and they don't automatically redistribute (I have to shut one down manually or by API).
@havoc said: Katapult.io and fly.io have me intrigued but haven't built anything on it yet.
I am disappointed by fly's failover algorithm. The servers seem to go down more often than they should, which results in a "black spot" until BGP updates and a new instance starts. Then I end up with two servers in a single region, and they don't automatically redistribute (I have to shut one down manually or by API).
Thanks for the heads up. Was looking into fly cause they offer free databases
Also just realised my katapult comment might not be useful to OP either...unsure whether their cores are dedicated
@havoc said: Katapult.io and fly.io have me intrigued but haven't built anything on it yet.
I am disappointed by fly's failover algorithm. The servers seem to go down more often than they should, which results in a "black spot" until BGP updates and a new instance starts. Then I end up with two servers in a single region, and they don't automatically redistribute (I have to shut one down manually or by API).
Thanks for the heads up. Was looking into fly cause they offer free databases
Also just realised my katapult comment might not be useful to OP either...unsure whether their cores are dedicated
I actually do use fly for DNS, and it has positives - anycast for free/cheap, and the ability to hook containers up via Wireguard has proved useful. For certain use cases (like when you've got some kind of failover or retry logic in the client) it is good.
Actually I just checked now and it looks like my containers have been up since October, so maybe the reliability has improved.
@havoc said: Katapult.io and fly.io have me intrigued but haven't built anything on it yet.
I am disappointed by fly's failover algorithm. The servers seem to go down more often than they should, which results in a "black spot" until BGP updates and a new instance starts. Then I end up with two servers in a single region, and they don't automatically redistribute (I have to shut one down manually or by API).
Thanks for the heads up. Was looking into fly cause they offer free databases
Also just realised my katapult comment might not be useful to OP either...unsure whether their cores are dedicated
I actually do use fly for DNS, and it has positives - anycast for free/cheap, and the ability to hook containers up via Wireguard has proved useful. For certain use cases (like when you've got some kind of failover or retry logic in the client) it is good.
Actually I just checked now and it looks like my containers have been up since October, so maybe the reliability has improved.
I'm considering building something for production on fly, how has been your experience with regards to reliability of containers running in a single region (as opposed to multi region)?
@havoc said: Katapult.io and fly.io have me intrigued but haven't built anything on it yet.
I am disappointed by fly's failover algorithm. The servers seem to go down more often than they should, which results in a "black spot" until BGP updates and a new instance starts. Then I end up with two servers in a single region, and they don't automatically redistribute (I have to shut one down manually or by API).
Thanks for the heads up. Was looking into fly cause they offer free databases
Also just realised my katapult comment might not be useful to OP either...unsure whether their cores are dedicated
I actually do use fly for DNS, and it has positives - anycast for free/cheap, and the ability to hook containers up via Wireguard has proved useful. For certain use cases (like when you've got some kind of failover or retry logic in the client) it is good.
Actually I just checked now and it looks like my containers have been up since October, so maybe the reliability has improved.
I'm considering building something for production on fly, how has been your experience with regards to reliability of containers running in a single region (as opposed to multi region)?
Hmm good question. It seems to vary a bit by location. In general I'd say "reliable" but not "rock solid". Probably in the Vultr type range of reliability. They have a "health check" system where if the container doesn't respond then it is shut down and redeployed. It can be configured in the toml file. I suspect this causes some false positives (i.e. the underlying containers are actually pretty good) and for a single region it would be worth trying to just shut off their health checks and do your own management using the API.
Comments
For my critical stuff, the truthful answer is that I use Azure.
For other production-quality stuff, I've had particularly good reliability with Avoro, UltraVPS.eu and Inception (at least in Europe).
If you configure LunaNode block storage as the boot volume, that can be pretty good because they reattach to a new node in the event of a problem. There can be some tricks with this (like avoiding UUIDs in the network config) but quite reliable.
@NDTN
Want free vps ? https://microlxc.net
https://lowendtalk.com/discussion/177258/netcup-rs-2000-g9-wsv22-16gb-ram-for-11-20-month#latest
Not sure if NVMe, but disk is quite fast (check yabs), and personally haven't had any issue with them
i would take a look at avoro (now owned by php-friends)
I bench YABS 24/7/365 unless it's a leap year.
Leaseweb
Vultr
It's SSD. Do they accept credit card (I don't like to pay via Paypal)
Is their uptime high? My priority is uptime. Cloudvider is one of the best uptime providers I tried, im looking for alternative offers.
I'm actually not sure with that
Always paid with PP
Past 12 months on my Avoro server.
i had one for 2 or 3 years, apart from planned migration (xeon gold to epyc) and planned network upgrade, there was one or 2 network downtime that was unplanned. machine uptime to my knowledge was 99.999% (really gut feel, not monitored)
I bench YABS 24/7/365 unless it's a leap year.
yes this. lol.
and the best thing? prepaid.
I bench YABS 24/7/365 unless it's a leap year.
What's the name of the monitoring software?
naemon (with 4 distributed monitoring workers)
AWS
However, given that your budget (in USD?) is
and NVMe is a must, based on what I read (can't speak from my own experience, sry
), I'd probably choose either of the following:
Contribute your idling VPS/dedi (link), Android (link) or iOS (link) devices to medical research
I went back and looked at my own stats for providers with > 6 months monitoring history.
The following had 100% uptime in the past 6 months: Avoro (Frankfurt), UltraVPS.eu (Düsseldorf), Inception (Amsterdam), GreenCloud (Dallas, Chicago), BuyVM (Las Vegas), HostHatch (Amsterdam), vServer.site (Düsseldorf), VirMach (various)
It is probably worth noting that the interval is 5 mins, so short outages may not be captured (seems like BuyVM is a bit susceptible to that). However, some of these are also monitored with Hetrix at a 1 minute interval and the stats there were generally in very close agreement.
Providers can differ a lot by location. HostHatch in Los Angeles is under 99% (poor), Inception in Phoenix is lower than in Europe (maybe 99.95% range), and UltraVPS.eu in Los Angeles had an outage in the past week which took them under 99.9%.
Mission critical stuff is on GCP, Cloudflare (workers/KV) and bunnyCDN. Generally try to engineer things to not use a VM cause GCP is expensive for that. So more of a blend of CF workers, cloud functions etc - think serverless vibes.
Katapult.io and fly.io have me intrigued but haven't built anything on it yet.
I'd probably do vultr because they seem to do well in comparisons, or katapult if feeling adventurous.
The LES servers I've got (couple of Nexus ones) are mostly testing rather than prod. But they seem good on uptime too
I am disappointed by fly's failover algorithm. The servers seem to go down more often than they should, which results in a "black spot" until BGP updates and a new instance starts. Then I end up with two servers in a single region, and they don't automatically redistribute (I have to shut one down manually or by API).
liteserver, hetzner
Amadex.com Domainer + IT Supporter | Brbljaona Balkan Chat Website | ICT Jobs Croatia
Thanks for the heads up. Was looking into fly cause they offer free databases
Also just realised my katapult comment might not be useful to OP either...unsure whether their cores are dedicated
I actually do use fly for DNS, and it has positives - anycast for free/cheap, and the ability to hook containers up via Wireguard has proved useful. For certain use cases (like when you've got some kind of failover or retry logic in the client) it is good.
Actually I just checked now and it looks like my containers have been up since October, so maybe the reliability has improved.
I'm considering building something for production on fly, how has been your experience with regards to reliability of containers running in a single region (as opposed to multi region)?
Hmm good question. It seems to vary a bit by location. In general I'd say "reliable" but not "rock solid". Probably in the Vultr type range of reliability. They have a "health check" system where if the container doesn't respond then it is shut down and redeployed. It can be configured in the toml file. I suspect this causes some false positives (i.e. the underlying containers are actually pretty good) and for a single region it would be worth trying to just shut off their health checks and do your own management using the API.