问题

您可以承受多少个节点失败？（4node cluster -rf2）

  • 2020年10月28日
  • 4个答复
  • 641视图
C
徽章

你好。我对RF2功能有疑问。

目前，它由4个节点和RF2作为群集组成。

据我所知，RF2有1个节点失败。

如果是这样，数据弹性状态不应该失败吗？

但是，尽管一个节点失败了，但数据弹性状态仍然可以。

实际上，当前群集中的CPU，内存和存储使用率很低。

目前，只有3个节点为Active CPU为20％的存储器为40％存储为18.5％。

为什么数据弹性状态可以？

即使目前关闭了一个节点，该服务也有任何问题吗？

还是群集会死亡，如果有1个节点死亡？

如果您不死，我想知道为什么。

\u00a0<\/p>

Hello. I have a question about the RF2 function.<\/p>

Currently, it is composed of 4 nodes and RF2 as a cluster.<\/p>

As far as I know, RF2 has 1 node fail.<\/p>

If so, shouldn't the Data Resiliency Status drop to Fail?<\/p>

However, although one node has failed, the Data Resiliency Status is still OK.<\/p>

In fact, the CPU, Memory, and Storage usage in the current cluster is low.<\/p>

Currently only 3 nodes are active CPU is 20% Memory is 40% Storage is 18.5%.<\/p>

Why is the Data Resiliency Status Ok?<\/p>

Is there any problem with the service even if one more node is currently turned off?<\/p>

Or will the cluster die if 1 more Node dies?<\/p>

If you don't die, I want to know why.<\/p>","quoteUsername":"cubensys","translations":{"Common":{"like":"Like","unlike":"Unlike"},"Forum":{"Quote":"Quote","Share":"Share"}}}">

4个答复

UPX
UserLevel 2
徽章 +4

你好 @cubensys

解释非常简单。
Nutanix不仅基于最小节点数量（群集为3），而且基于资源的可用性。

这对您而言是什么意思？

您有4个带有RF2的节点，当一个节点失败时，AOS立即开始复制数据以返回稳定的RF2状态。由于您有4个节点，并且可能有足够的资源可以使工作负载保持活跃，因此系统仅用3个节点返回稳定状态。

如果您的群集处于3个节点或3个活跃节点上的资源不足以使工作负载完全运行，那么这显然不会发生。

现在，将数据弹性处于确定状态，您可以失去另一个节点，然后您将面临关键状态。

希望这可以帮助

准备工作是出色工作的起点！
Hi @cubensys<\/user-mention>\u00a0<\/p>

The explanation is fairly simple.
Nutanix bases the resilience factor not only on the minimum number of nodes (which for a cluster is 3) but also on the availability of resources.<\/p>

What does this mean in your case?<\/p>

You have 4 nodes with RF2, when one of the nodes fails AOS\u00a0immediately starts replicating the data to return to a stable RF2 state. Since you have 4 nodes and probably enough resources to keep the workloads active, the system returns to a stable state with only 3 nodes.<\/p>

This obviously would not have happened if your cluster had been at 3 nodes or if the resources left on the 3 active nodes were not enough to keep the workloads fully operational.<\/p>

Now, with your data resiliency in OK state, you can loose another node and then you will face\u00a0a critical status.<\/p>

Hope this helps<\/p>","quoteUsername":"UPX","translations":{"Common":{"like":"Like","unlike":"Unlike"},"Forum":{"Quote":"Quote","Share":"Share"}}}">

C
徽章 
谢谢您的回答。资源可用性？
Thank you for answer.Then, I understood that the cluster is in a stable state with 3 nodes.If one fails in three nodes, the Data Resiliency Status will be in a critical state,\u00a0but can it be determined that the service is maintained due to resource availability?<\/pre>","quoteUsername":"cubensys","translations":{"Common":{"like":"Like","unlike":"Unlike"},"Forum":{"Quote":"Quote","Share":"Share"}}}">
UPX
UserLevel 2
徽章 +4 
谢谢您的回答。资源可用性？

不确定我是否理解您的问题，但是通常，如果您对N+1的3个节点群集大小，当一个节点失败时，临界状态仅意味着您无法放松其他组件，但所有工作负载都启动并运行。否则，系统仅维护运行的工作负载，而其他资源则停止。

您可以确定您的群集将用于3种不同的设置的HA类型

最好的努力（系统将根据资源消耗自动决定）

HA（系统将保留每个节点的专用内存，以实现HA的目的）

专用节点（系统将整个节点荣誉用作备用）此设置已弃用

您会在此处找到有关HA如何工作的更多有用信息

https://portal.nutanix.com/page/documents/kbs/details?targetId=ka00e0000000000liquca4

准备工作是出色工作的起点！
\t
Thank you for answer.Then, I understood that the cluster is in a stable state with 3 nodes.If one fails in three nodes, the Data Resiliency Status will be in a critical state,\u00a0but can it be determined that the service is maintained due to resource availability?<\/pre>\t<\/content-quote>
Not sure if i understand your question but usually, if you size\u00a0a 3 nodes cluster for N+1, when a node fails the critical state simply means you can not loose other components\u00a0but all the workloads are up and running. Otherwise the system maintain running only the workload there are resources for, the others are stopped.\u00a0<\/p>
You can be sure about the type\u00a0of HA\u00a0your cluster will use with 3 different settings<\/p>
Best Effort (the system will decide automatically based on resource consumption)<\/p>
HA (The system will reserve a dedicate amount of memory per node for HA purpose)<\/p>
Dedicated Node (An entire node wolud be used by the system as spare) this setting\u00a0is deprecated<\/u><\/strong><\/span><\/p>
You will find more useful\u00a0information about how HA works here<\/p>
https:\/\/portal.nutanix.com\/page\/documents\/kbs\/details?targetId=kA00e000000LIQUCA4<\/a><\/p>
\u00a0<\/p>
\u00a0<\/p>
\u00a0<\/p>","quoteUsername":"UPX","translations":{"Common":{"like":"Like","unlike":"Unlike"},"Forum":{"Quote":"Quote","Share":"Share"}}}">
阿罗纳
秩
UserLevel 6
徽章 +5

嗨，Cubensys，

与其从节点角度观察任务，而要从数据观点的副本数量查看。

您有4个节点。在RF-2群集中，您有2个数据副本。您会失去一个节点，这意味着您将丢失数据的第二副本的一部分。预期的初始数据弹性状态失败了，因为如果您丢失了另一个节点，则有可能在该节点中丢失相同数据的另一个副本。

但是，数据的第二副本仍然可以通过复制现有数据来恢复。复制过程完成后，群集数据弹性状态返回到确定。

Hi cubensys,<\/p>

\u00a0<\/p>

Instead of looking at the task from the number of nodes perspective, look at it from the number of copies of data point of view.<\/p>

You have 4 nodes. In an RF-2 cluster you have 2 copies of data. You loose one node which means you lose part of the second copy of the data. The expected initial Data Resiliency state is Fail because if you lose another node then it is possible that with that node you lose the other copy of the same data.<\/p>

However the second copy of data can still be recovered by replicating the existing data. Once the copying process is complete the cluster Data Resiliency status returns to OK.<\/p>","quoteUsername":"Alona","translations":{"Common":{"like":"Like","unlike":"Unlike"},"Forum":{"Quote":"Quote","Share":"Share"}}}">

回复


Baidu