VULTR NJ location: Partial Power failure.
-
All very good points to consider. And yes, if you have a snapshot, you can go to another location and restore it, or as mentioned, fail over.
And no, it’s impossible to say that you have one hundred percent up time. Even on a on prem server will have down time from time to time.
Yes, a fail over is a better option to ‘count on’
I just found it quite interesting that it occurred
-
@gjacobse said in VULTR NJ location: Partial Power failure.:
What do you do when the hosted service has a power failure which brings your instance down?
I wasn’t affected, but I’d be curious to see how the system is designed - over all - and see what the break in redundancy was.
Since vultr is using consumer CPUs to bring the cost down I bet their infrastructure is at the same level.
Power outages never brings the servers down if the servers sit in a tier 3 (or better) datacenter.
What you do is pay more and host you workloads elsewhere or accept that the service provided is in relationship to the price.
-
Wow.
I read his opening question with the emphasis on "you". I didn't read anything like @gjacobse wanting to lay blame or hear about all the woulda, coulda, shoulda scenarios. He didn't even say if his instance was critical.I wonder if he was just looking for conversation about "how the system is designed - over all - and see what the break in redundancy was".
-
Out of interest do Vultr / Azure / AWS / etc offer HA for instances between different locations?
-
@hobbit666 said in VULTR NJ location: Partial Power failure.:
Out of interest do Vultr / Azure / AWS / etc offer HA for instances between different locations?
No. Not even their "Load Balancer" thing does that.
https://www.vultr.com/products/load-balancers/
But VPS provider != Cloud provider
-
@hobbit666 said in VULTR NJ location: Partial Power failure.:
Out of interest do Vultr / Azure / AWS / etc offer HA for instances between different locations?
You are mixing things up here.
Vultr is not a cloud provider, they are a VPS provider
AWS and Azure are cloud providers. They both offer regional redundancy/recovery options.
-
@JaredBusch said in VULTR NJ location: Partial Power failure.:
@hobbit666 said in VULTR NJ location: Partial Power failure.:
Out of interest do Vultr / Azure / AWS / etc offer HA for instances between different locations?
You are mixing things up here.
Vultr is not a cloud provider, they are a VPS provider
AWS and Azure are cloud providers. They both offer regional redundancy/recovery options.
Then yes i am
thought they were all the same lol. -
It looks like Vultr have their servers in colo at the QTS Piscataway NJ datacenter.
https://www.qtsdatacenters.com/data-centers/piscatawayThe datacenter looks to be the real deal so "a partial power outage to a subset of servers" is probably just a rack PDU failure or possibly overloading of a rack so circuit breakers cut the power.
-
@Pete-S said in VULTR NJ location: Partial Power failure.:
It looks like Vultr have their servers in colo at the QTS Piscataway NJ datacenter.
https://www.qtsdatacenters.com/data-centers/piscatawayThe datacenter looks to be the real deal so "a partial power outage to a subset of servers" is probably just a rack PDU failure or possibly overloading of a rack so circuit breakers cut the power.
Assuming enterprise gear all servers have dual power supplies and each rack has two PDUs. Every equipment draws power from both PDUs. So each PDU carries half the power of the rack.
If one PDU breaks the other PDU has to carry all power. Now it will get twice the current and if that is too much (if you didn't pay attention to the power draw), the circuit breakers to the working PDU will cut the power to the entire rack and all servers goes down.
One rack with about 40 servers will probably host a good amount of VPSs.
-
@gjacobse said in VULTR NJ location: Partial Power failure.:
I wasn’t affected, but I’d be curious to see how the system is designed - over all - and see what the break in redundancy was.
It's cloud, redundancy is meant to come from you, not the DC.
-
@gjacobse said in VULTR NJ location: Partial Power failure.:
What do you do when the hosted service has a power failure which brings your instance down?
Generally nothing. It's rare and generally just a few minutes, it almost never causes actual impact.
-
@gjacobse said in VULTR NJ location: Partial Power failure.:
Even on a on prem server will have down time from time to time.
On prem typically as the most.
-
@Pete-S said in VULTR NJ location: Partial Power failure.:
Since vultr is using consumer CPUs to bring the cost down I bet their infrastructure is at the same level.
We see fewer unplanned outages than with AWS. How do you feel about Amazon based on that?
-
@JaredBusch said in VULTR NJ location: Partial Power failure.:
Vultr is not a cloud provider, they are a VPS provider
Vultr is absolutely cloud. Cloud presented with a VPN interface. It's still cloud in every way. Just like Digital Ocean or Linode or AWS (via Lightsail.)
-
@Pete-S said in VULTR NJ location: Partial Power failure.:
Since vultr is using consumer CPUs to bring the cost down I bet their infrastructure is at the same level.
Consumer CPUs would normally do the opposite, that's why enterprises use bigger processors, to get more bang for the buck in a smaller space.
But how did you determine that? I just looked at my servers and their specs only match server class Xeon processors (Skylake 16MB L3) so not sure how it is possible for them to be consumer.
-
@scottalanmiller said in VULTR NJ location: Partial Power failure.:
@Pete-S said in VULTR NJ location: Partial Power failure.:
Since vultr is using consumer CPUs to bring the cost down I bet their infrastructure is at the same level.
We see fewer unplanned outages than with AWS. How do you feel about Amazon based on that?
What do you have running in AWS for which you have outages?
-
@scottalanmiller said in VULTR NJ location: Partial Power failure.:
But how did you determine that? I just looked at my servers and their specs only match server class Xeon processors (Skylake 16MB L3) so not sure how it is possible for them to be consumer.
I don't remember exactly. A couple of years ago you could determine that there was actually no xeon of that architecture that would fit with the GHz. Since they are obscuring the actual CPU in their linux kernel you can't read the model number outright. I going from memory here, but that was the gist of it.
-
This post is deleted!
