Should I bother to learn Windows Storage Spaces and what about Glances export?
-
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
Contrary to the naysayers we've done quite well and so have many others including Tier 1.
The naysayers NEVER suggest that the issues will happen to everyone, that's not how risk works - so there is no contrary here. This exactly the "it works for me" issue that had already been mentioned. Stating that it "does work for some people" highlights a misunderstanding of risk.
https://mangolassi.it/topic/14750/ask-your-mom-to-explain-risk
If you understand how risk works, and understand that for production storage we need things like seven nines durability or higher, than you know that even knowing hundreds of companies where it works tells us essentially nothing. And it is Microsoft themselves who said it wasn't really ready and wasn't being used by a statistically useful number of production installs.
We've had this discussion before on many different topics and at this point I am of the opinion that we shall need to agree to disagree.
We have six to seven years of deploying Storage Spaces, Hyper-V, and now Storage Spaces Direct clusters. I'd be more than happy to chat about the what/where/how of what we do. @Emad-R DM me.
No one is doubting that you've done it for a while, nor that you've had good results. What I'm trying to say is that even if SS was horribly dangerous, you still expect to find lots of people for whom it has worked. That you've had good experiences, even quite a few, over a long period of time, doesn't tell us that it's safe, or even imply such. It simply states that it is "capable of working." It doesn't make it bad, either. It just doesn't give us anything useful for statistics.
It's like that IPOD situation I dealt with yesterday. Sixteen years it "worked fine." Using math, we knew what they did was dangerous, but from observation you'd easily say "yeah, but it has protected us for SO long, how can it be bad?" Then, when it finally failed, it was clear that it was the IPOD design that caused the failure, exactly as it was most likely to do so. And we could show that in that specific scenario, skipping the IPOD would alone have protected them.
It's not that I'm disagreeing with your experience. It's that I'm stating how that experience applies to risk math.
-
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
-
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
Server maintenance must be performed in any case. An IPOD or a single server still has this requirement. But with an IPOD maintenance could introduce issues that you wouldn't see with a single server.
-
Maintenance is just that, maintenance. But IPODs introduce complexity for often little to no gain.
A single host while still requiring maintenance has much fewer pieces that need to be maintained. Thus reducing the risk that performing maintenance causes an issue that would be unexpected.
IE A firmware update on your Switch breaks iSCSI connections.
-
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
They were SO scared of the system's complexity that they avoided maintaining it. And the system did NOT provide full true redundancy, so had to be taken offline for some maintenance anyway.
-
@DustinB3403 said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
Server maintenance must be performed in any case. An IPOD or a single server still has this requirement. But with an IPOD maintenance could introduce issues that you wouldn't see with a single server.
The IPOD also required much more maintenance.
-
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
We deploy traditional RAID in standalone settings not Storage Spaces.
-
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
-
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
-
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
Only if you are looking for HA/Failover. On a single host with local storage you wouldn't need any of this.
-
@DustinB3403 said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
Only if you are looking for HA/Failover. On a single host with local storage you wouldn't need any of this.
That is true. Most cars have 4 tires. Also true.
-
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
This is the 2-node shared SAS Hyper-V/Storage Spaces cluster mentioned above that runs a 15-18 seat accounting firm.
There are two types of virtual disks set up on Storage Spaces. One with a 64KB interleave with the storage stack similarly configured while the other is the standard 256KB interleave with the defaults for storage stack. There are six to eight server based virtual machines and at least two or three desktop virtual machines running on the cluster at any given time.
EDIT: There are multiple virtual disks set up as Cluster Shared Volumes.
-
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
This is the 2-node shared SAS Hyper-V/Storage Spaces cluster mentioned above that runs a 15-18 seat accounting firm.
There are two types of virtual disks set up on Storage Spaces. One with a 64KB interleave with the storage stack similarly configured while the other is the standard 256KB interleave with the defaults for storage stack. There are six to eight server based virtual machines and at least two or three desktop virtual machines running on the cluster at any given time.
EDIT: There are multiple virtual disks set up as Cluster Shared Volumes.
@PhlipElder cool cool. . . so what happens if that dataon unit fails to the 9's?
Your client would be dead in the water, no?
-
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
This is the 2-node shared SAS Hyper-V/Storage Spaces cluster mentioned above that runs a 15-18 seat accounting firm.
There are two types of virtual disks set up on Storage Spaces. One with a 64KB interleave with the storage stack similarly configured while the other is the standard 256KB interleave with the defaults for storage stack. There are six to eight server based virtual machines and at least two or three desktop virtual machines running on the cluster at any given time.
EDIT: There are multiple virtual disks set up as Cluster Shared Volumes.
Oh SWEET! DataOn. Yeah I wouldn't use anything but that for the purpose.
-
What you are showing us is nothing more than an IPOD with Hyper-V and S2D.
-
@DustinB3403 said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
What you are showing us is nothing more than an IPOD with Hyper-V and S2D.
It seems like he has disk redundancy, but not DAS redundancy.... if I understand the post correctly.
@PhlipElder You have two DAS boxes, are they redundant or stacked? It seems as if they are stacked, making what Dustin said true.
-
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
SOFS does that. But you can use a large range of RAID or RAIN systems to handle it. Gluster, CEPH, Starwind, DRBD, HAST, etc.
-
@DustinB3403 said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
Only if you are looking for HA/Failover. On a single host with local storage you wouldn't need any of this.
He already put in the "to get around" piece.
-
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@DustinB3403 said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
Only if you are looking for HA/Failover. On a single host with local storage you wouldn't need any of this.
That is true. Most cars have 4 tires. Also true.
The difference being that most cars should have four tires, most workloads should not have zero downtime maintenance. Some workloads, but not the majority.
-
@DustinB3403 said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@PhlipElder said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@Obsolesce said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
@scottalanmiller said in Should I bother to learn Windows Storage Spaces and what about Glances export?:
It's like that IPOD situation I dealt with yesterday.
What about all of the server maintenance able to be done without having any down time? Or didn't they use it for that, strictly redundancy?
In a cluster setting (SOFS) this is a moot point since nodes can be patched and rebooted without any downtime.
Including the SOFS nodes, you mean. That's the important part. It fixes the single maintenance point of the SAN.
You'd need S2D (or similar tech, like SW vSAN) to get around the single maintenance point of SAN / DAS.
This is the 2-node shared SAS Hyper-V/Storage Spaces cluster mentioned above that runs a 15-18 seat accounting firm.
There are two types of virtual disks set up on Storage Spaces. One with a 64KB interleave with the storage stack similarly configured while the other is the standard 256KB interleave with the defaults for storage stack. There are six to eight server based virtual machines and at least two or three desktop virtual machines running on the cluster at any given time.
EDIT: There are multiple virtual disks set up as Cluster Shared Volumes.
@PhlipElder cool cool. . . so what happens if that dataon unit fails to the 9's?
Your client would be dead in the water, no?
The unit is fully redundant all the way through to the disk. If we have a complete system failure we have Veeam and the ability to spin the VMs up on short order.
