Posts Tagged ‘failover’

vRealize Automation Disaster Recovery

January 14, 2018

Introduction

VMware has invested a lot of time and effort in vRealize Automation high availability. For medium and large deployment scenarios VMware recommends using a load balancer (Citrix, F5, NSX) to distribute traffic between vRA appliance and infrastructure components, as well as database clustering (such as MS SQL availability groups) for database high availability. Additionally, in vRA 7.3 VMware added support for automatic failover of vRA appliance’s embedded PostgreSQL database, which was a manual process prior to that.

There is a clear distinction, however, between high availability and disaster recovery. Generally speaking, HA covers redundancy within the site and is not intended to protect from full site failure. Site Recovery Manager (or another replication product) is required to protect vRA in a DR scenario, which is described in more detail in the following document:

In my opinion, there are two important aspects that are missing from the aforementioned document, which I want to cover in this blog post: restoring VM UUIDs and changing vRA IP address. I will cover them in the order that these tasks would usually be performed if you were to fail over vRA to DR:

  1. Exporting VM UUIDs
  2. Changing IP addresses
  3. Importing VM UUIDs

I will also only touch on how to change VM reservations. Which is also an important step, but very well covered in VMware documentation already.

Note: this blog post does not provide configuration guidelines for VM replication software, such as Site Recovery Manager, Zerto or RecoverPoint and is focused only on DR aspects related to vRA itself. Refer to official documentation of corresponding products to determine how to set up VM replication to your disaster recovery site.

Exporting VM UUIDs

VMware uses two UUIDs to identify a VM. BIOS UUID (uuid.bios in .vmx file) was the original VM identifier implemented to identify a VM and is derived from the hardware VM is provisioned on. But it’s not unique. If VM is cloned, the clone will have the same BIOS UUID. So the second identifier was introduced called Instance UUID (vc.uuid in .vmx file), which is generated by vCenter and is unique within a single vCenter (two VMs in different vCenters can have the same Instance UUID).

When VMs are failed over, Instance UUIDs change. Compare VirtualMachine.Admin.AgentID (Instance UUID) and VirtualMachine.Admin.UUID (BIOS UUID) on original and failed over VMs.

Why does this matter? Because vRA uses Instance UUIDs to keep track of managed VMs.  If Instance UUIDs change, vRA will show the corresponding VMs as missing under Infrastructure > Managed Machines. And you won’t be able to manage them.

So it’s important to export VM Instance UUIDs before failover, which can then be used to restore the original values. This is how you can get the Instance UUID of a given VM using PowerCLI:

> (Get-VM vm_name).extensiondata.config.InstanceUUID

Here, on my GitHub page, you can find a script that I have put together to export Instance UUIDs of all VMs in CSV format.

Changing IP addresses

Once you’ve saved the Instance UUIDs, you can move on to failover. vRA components should be started in the following order:

  1. MS SQL database
  2. vRA appliance
  3. IaaS server

If network subnets, that all components are connected to, are stretched between two sites, when VMs are brought up at DR, there are no additional reconfiguration required. But usually it’s not the case and servers need to be re-IP’ed. IaaS server network setting are changed the same as on any other Windows server machine.

vRealize Appliance network settings are changed in vRA appliance management interface, that can be accessed at https://vra-appliance-hostname:5480, under Network > Address tab. The problem is, if IP addresses change at DR, it will be challenging to reach vRA appliance over the network. To work around that, connect to vRA VM console and run the following script from CLI to change appliance’s network settings:

# /opt/vmware/share/vami/vami_config_net

Don’t forget to update the DNS record for vRA appliance in DNS. For IaaS server it’s not needed, as long as you allow Dynamic DNS (DDNS) updates.

Importing VM UUIDs

After the failover all of your VMs will have missing status in vRA. To make vRA recognize failed over VMs you will need to revert Instance UUIDs back to the original values. In PowerCLI this can be done in the following way:

> $spec = New-Object VMware.Vim.VirtualMachineConfigSpec
> $spec.instanceUuid = ’52da9b14-0060-dc51-4733-3b01e912edd2′
> $vm = Get-VM -Name vm_name
> $vm.Extensiondata.ReconfigVM_Task($spec)

I’ve written another script, that will perform this task for you, which you can find on my GitHub page.

You will need two files to make the script work. The vm_vc_uuids.csv file you generated before, with the list of original VM Instance UUIDs. As well as the list of missing VMs in CSV format, that you can export from vRA after the failover on the Infrastructure > Managed Machines page:

This is an example of the script command line options and the output:

You will need to run an inventory data collection from the Infrastructure > Compute Resources > Compute Resources page. vRA will discover VMs and update their status to “On”.

Updating reservations

If you try to run any Day 2 operation on a VM with the old reservation in place, you will get an error similar to this:

Error processing [Shutdown], error details:
Error getting property ‘runtime.powerState’ from managed object (null)
Inner Exception: Object reference not set to an instance of an object.

To manually update VM reservation, on Infrastructure > Managed Machines page hover over the VM and select Change Reservation:

This process is obviously not scalable, as it can take hours, if you have hundreds of VMs. VMware offers an alternative solution that lets you update all VMs by using Bulk Import feature available from Infrastructure > Administration > Bulk Imports. The idea is that you can export all VM configuration details in a CSV file, update compute and storage reservation columns and import back to vRA. vRealize Suite 7.0 Disaster Recovery by Using Site Recovery Manager 6.1 gives very detailed instruction on how to do that in “Bulk Import, Update, or Migrate Virtual Machines” section.

Conclusion

I hope this blog post helped to cover some gaps in VMware documentation. If you have any questions or comments, as always, feel free to leave them in the comments sections below.

References

Advertisements

Dell Compellent is not an ALUA Storage Array

May 16, 2016

dell_compellentDell Compellent is Dell’s flagship storage array which competes in the market with such rivals as EMC VNX and NetApp FAS. All these products have slightly different storage architectures. In this blog post I want to discuss what distinguishes Dell Compellent from the aforementioned arrays when it comes to multipathing and failover. This may help you make right decisions when designing and installing a solution based on Dell Compellent in your production environment.

Compellent Array Architecture

In one of my previous posts I showed how Compellent LUNs on vSphere ESXi hosts are claimed by VMW_SATP_DEFAULT_AA instead of VMW_SATP_ALUA SATP, which is the default for all ALUA arrays. This happens because Compellent is not actually an ALUA array and doesn’t have the tpgs_on option enabled. Let’s digress for a minute and talk about what the tpgs_on option actually is.

For a storage array to be claimed by VMW_SATP_ALUA it has to have the tpgs_on option enabled, as indicated by the corresponding SATP claim rule:

# esxcli storage nmp satp rule list

Name                 Transport  Claim Options Description
-------------------  ---------  ------------- -----------------------------------
VMW_SATP_ALUA                   tpgs_on       Any array with ALUA support

This is how Target Port Groups (TPG) are defined in section 5.8.2.1 Introduction to asymmetric logical unit access of SCSI Primary Commands – 3 (SPC-3) standard:

A target port group is defined as a set of target ports that are in the same target port asymmetric access state at all times. A target port group asymmetric access state is defined as the target port asymmetric access state common to the set of target ports in a target port group. The grouping of target ports is vendor specific.

This has to do with how ports on storage controllers are grouped. On an ALUA array even though a LUN can be accessed through either of the controllers, paths only to one of them (controller which owns the LUN) are Active Optimized (AO) and paths to the other controller (non-owner) are Active Non-Optimized (ANO).

Compellent does not present LUNs through the non-owning controller. You can easily see that if you go to the LUN properties. In this example we have four iSCSI ports connected (two per controller) on the Compellent side, but we can see only two paths, which are the paths from the owning controller.

compellent_psp

If Compellent presents each particular LUN only through one controller, then how does it implement failover? Compellent uses a concept of fault domains and control ports to handle LUN failover between controllers.

Compellent Fault Domains

This is Dell’s definition of a Fault Domain:

Fault domains group front-end ports that are connected to the same Fibre Channel fabric or Ethernet network. Ports that belong to the same fault domain can fail over to each other because they have the same connectivity.

So depending on how you decided to go about your iSCSI network configuration you can have one iSCSI subnet / one fault domain / one control port or two iSCSI subnets / two fault domains / two control ports. Either of the designs work fine, this is really is just a matter of preference.

You can think of a Control Port as a Virtual IP (VIP) for the particular iSCSI subnet. When you’re setting up iSCSI connectivity to a Compellent, you specify Control Ports IPs in Dynamic Discovery section of the iSCSI adapter properties. Which then redirects the traffic to the actual controller IPs.

If you go to the Storage Center GUI you will see that Compellent also creates one virtual port for every iSCSI physical port. This is what’s called a Virtual Port Mode and is recommended instead of a Physical Port Mode, which is the default setting during the array initialization.

Failover scenarios

Now that we now what fault domains are, let’s talk about the different failover scenarios. Failover can happen on either a port level when you have a transceiver / cable failure or a controller level, when the whole controller goes down or is rebooted. Let’s discuss all of these scenarios and their variations one by one.

1. One Port Failed / One Fault Domain

If you use one iSCSI subnet and hence one fault domain, when you have a port failure, Compellent will move the failed port to the other port on the same controller within the same fault domain.

port_failed

In this example, 5000D31000B48B0E and 5000D31000B48B0D are physical ports and 5000D31000B48B1D and 5000D31000B48B1C are the corresponding virtual ports on the first controller. Physical port 5000D31000B48B0E fails. Since both ports on the controller are in the same fault domain, controller moves virtual port 5000D31000B48B1D from its original physical port 5000D31000B48B0E to the physical port 5000D31000B48B0D, which still has connection to network. In the background Compellent uses iSCSI redirect command on the Control Port to move the traffic to the new virtual port location.

2. One Port Failed / Two Fault Domains

Two fault domains scenario is slightly different as now on each controller there’s only one port in each fault domain. If any of the ports were to fail, controller would not fail over the port. Port is failed over only within the same controller/domain. Since there’s no second port in the same fault domain, the virtual port stays down.

port_failed_2

A distinction needs to be made between the physical and virtual ports here. Because from the physical perspective you lose one physical link in both One Fault Domain and Two Fault Domains scenarios. The only difference is, since in the latter case the virtual port is not moved, you’ll see one path down when you go to LUN properties on an ESXi host.

3. Two Ports Failed

This is the scenario which you have to be careful with. Compellent does not initiate a controller failover when all front-end ports on a controller fail. The end result – all LUNs owned by this controller become unavailable.

two_ports_failed_2

luns_down

This is the price Compellent pays for not supporting ALUA. However, such scenario is very unlikely to happen in a properly designed solution. If you have two redundant network switches and controllers are cross-connected to both of them, if one switch fails you lose only one link per controller and all LUNs stay accessible through the remaining links/switch.

4. Controller Failed / Rebooted

If the whole controller fails the ports are failed over in a similar fashion. But now, instead of moving ports within the controller, ports are moved across controllers and LUNs come across with them. You can see how all virtual ports have been failed over from the second (failed) to the first (survived) controller:

controller_failed

Once the second controller gets back online, you will need to rebalance the ports or in other words move them back to the original controller. This doesn’t happen automatically. Compellent will either show you a pop up window or you can do that by going to System > Setup > Multi-Controller > Rebalance Local Ports.

Conclusion

Dell Compellent is not an ALUA storage array and falls into the category of Active/Passive arrays from the LUN access perspective. Under such architecture both controller can service I/O, but each particular LUN can be accessed only through one controller. This is different from the ALUA arrays, where LUN can be accessed from both controllers, but paths are active optimized on the owning controller and active non-optimized on the non-owning controller.

From the end user perspective it does not make much of a difference. As we’ve seen, Compellent can handle failover on both port and controller levels. The only exception is, Compellent doesn’t failover a controller if it loses all front-end connectivity, but this issue can be easily avoided by properly designing iSCSI network and making sure that both controllers are connected to two upstream switches in a redundant fashion.

How Admission Control Really Works

May 2, 2016

confusionThere is a moment in every vSphere admin’s life when he faces vSphere Admission Control. Quite often this moment is not the most pleasant one. In one of my previous posts I talked about some of the common issues that Admission Control may cause and how to avoid them. And quite frankly Admission Control seems to do more harm than good in most vSphere environments.

Admission Control is a vSphere feature that is built to make sure that VMs with reservations can be restarted in a cluster if one of the cluster hosts fails. “Reservations” is the key word here. There is a common belief that Admission Control protects all other VMs as well, but that’s not true.

Let me go through all three vSphere Admission Control policies and explain why you’re better of disabling Admission Control altogether, as all of these policies give you little to no benefit.

Host failures cluster tolerates

This policy is the default when you deploy a vSphere cluster and policy which causes the most issues. “Host failures cluster tolerates” uses slots to determine if a VM is allowed to be powered on in a cluster. Depending on whether VM has CPU and memory reservations configured it can use one or more slots.

Slot Size

To determine the total number of slots for a cluster, Admission Control uses slot size. Slot size is either the default 32MHz and 128MB of RAM (for vSphere 6) or if you have VMs in the cluster configured with reservations, then the slot size will be calculated based on the maximum CPU/memory reservation. So say if you have 100 VMs, 98 of which have no reservations, one VM has 2 vCPUs and 8GB of memory reserved and another VM has 4 vCPUs and 4GB of memory reserved, then the slot size will jump from 32MHz / 128MB to 4 vCPUs / 8GB of memory. If you have 2.0 GHz CPUs on your hosts, the 4 vCPU reservation will be an equivalent of 8.0 GHz.

Total Number of Slots

Now that we know the slot size, which happens to be 8.0 GHz and 8GB of memory, we can calculate the total number of slots in the cluster. If you have 2 x 8 core CPUs and 256GB of RAM in each of 4 ESXi hosts, then your total amount of resources is 16 cores x 2.0 GHz x 4 hosts = 128 GHz and 256GB x 4 hosts = 1TB of RAM. If your slot size is 4 vCPUs and 8GB of RAM, you get 64 vCPUs / 4 vCPUs = 16 slots (you’ll get more for memory, but the least common denominator has to be used).

total_slots

Practical Use

Now if you configure to tolerate one host failure, you have to subtract four slots from the total number. Every VM, even if it doesn’t have reservations takes up one slot. And as a result you can power on maximum 12 VMs on your cluster. How does that sound?

Such incredibly restrictive behaviour is the reason why almost no one uses it in production. Unless it’s left there by default. You can manually change the slot size, but I have no knowledge of an approach one would use to determine the slot size. That’s the policy number one.

Percentage of cluster resources reserved as failover spare capacity

This is the second policy, which is commonly recommended by most to use instead of the restrictive “Host failures cluster tolerates”. This policy uses percentage-based instead of the slot-based admission.

It’s much more straightforward, you simply specify the percentage of resources you want to reserve. For example if you have four hosts in a cluster the common belief is that if you specify 25% of CPU and memory, they’ll be reserved to restart VMs in case one of the hosts fail. But it won’t. Here’s the reason why.

When calculating amount of free resources in a cluster, Admission Control takes into account only VM reservations and memory overhead. If you have no VMs with reservations in your cluster then HA will be showing close to 99% of free resources even if you’re running 200 VMs.

failover_capacity

For instance, if all of your VMs have 4 vCPUs and 8GB of RAM, then memory overhead would be 60.67MB per VM. For 300 VMs it’s roughly 18GB. If you have two VMs with reservations, say one VM with 2 vCPUs / 4GB of RAM and another VM with 4 vCPUs / 2GB of RAM, then you’ll need to add up your reservations as well.

So if we consider memory, it’s 18GB + 4GB + 2GB = 24GB. If you have the total of 1TB of RAM in your cluster, Admission Control will consider 97% of your memory resources being free.

For such approach to work you’d need to configure reservations on 100% of your VMs. Which obviously no one would do. So that’s the policy number two.

Specify failover hosts

This is the third policy, which typically is the least recommended, because it dedicates a host (or multiple hosts) specifically just for failover. You cannot run VMs on such hosts. If you try to vMotion a VM to it, you’ll get an error.

failover_host

In my opinion, this policy would actually be the most useful for reserving cluster resources. You want to have N+1 redundancy, then reserve it. This policy does exactly that.

Conclusion

When it comes to vSphere Admission Control, everyone knows that “Host failures cluster tolerates” policy uses slot-based admission and is better to be avoided.

There’s a common misconception, though, that “Percentage of cluster resources reserved as failover spare capacity” is more useful and can reserve CPU and memory capacity for host failover. But in reality it’ll let you run as many VMs as you want and utilize all of your cluster resources, except for the tiny amount of CPU and memory for a handful of VMs with reservations you may have in your environment.

If you want to reserve failover capacity in your cluster, either use “Specify failover hosts” policy or simply disable Admission Control and keep an eye on your cluster resource utilization manually (or using vROps) to make sure you always have room for growth.

Implications of Ignoring vSphere Admission Control

April 5, 2016

no-admissionHA Admission Control has historically been on of the lesser understood vSphere topics. It’s not intuitive how it works and what it does. As a result it’s left configured with default values in most vSphere environments. But default Admission Control setting are very restrictive and can often cause issues.

In this blog post I want to share the two most common issues with vSphere Admission Control and solutions to these issues.

Issue #1: Not being able to start a VM

Description

Probably the most common issue everyone encounters with Admission Control is when you suddenly cannot power on VMs any more. There are multiple reasons why that might happen, but most likely you’ve just configured a reservation on one of your VMs or deployed a VM from an OVA template with a pre-configured reservation. This has triggered a change in Admission Control slot size and based on the new slot size you no longer have enough slots to satisfy failover requirements.

As a result you get the following alarm in vCenter: “Insufficient vSphere HA failover resources”. And when you try to create and boot a new VM you get: “Insufficient resources to satisfy configured failover level for vSphere HA”.

admission_error

Cause

So what exactly has happened here. In my example a new VM with 4GHz of CPU and 4GB of RAM was deployed. Admission Control was set to its default “Host Failures Cluster Tolerates” policy. This policy uses slot sizes. Total amount of resources in the cluster is divided by the slot size (4GHz and 4GB in the above case) and then each VM (even if it doesn’t have a reservation) uses at least 1 slot. Once you configure a VM reservation, depending on the number of VMs in your cluster more often than not you get all slots being used straight away. As you can see based on the calculations I have 91 slots in the cluster, which have instantly been used by 165 running VMs.

slot_calculations

Solution

You can control the slot size manually and make it much smaller, such as 1GHz and 1GB of RAM. That way you’d have much more slots. The VM from my previous example would use four slots. And all other VMs which have no reservations would use less slots in total, because of a smaller slot size. But this process is manual and prone to error.

The better solution is to use “Percentage of Cluster Resources” policy, which is recommended for most environments. We’ll go over the main differences between the three available Admission Control policies after we discuss the second issue.

Issue #2: Not being able to enter Maintenance Mode

Description

It might be a corner case, but I still see it quite often. It’s when you have two hosts in a cluster (such as ROBO, DR or just a small environment) and try to put one host into maintenance mode.

The first issue you will encounter is that VMs are not automatically vMotion’ed to other hosts using DRS. You have to evacuate VMs manually.

And then once you move all VMs to the other host and put it into maintenance mode, you again can no longer power on VMs and get the same error: “Insufficient resources to satisfy configured failover level for vSphere HA”.

poweron_fail

Cause

This happens because disconnected hosts and hosts in maintenance mode are not used in Admission Control calculations. And one host is obviously not enough for failover, because if it fails, there are no other hosts to fail over to.

Solution

If you got caught up in such situation you can temporarily disable Admission Control all together until you finish maintenance. This is the reason why it’s often recommended to have at least 3 hosts in a cluster, but it can not always be justified if you have just a handful of VMs.

Alternatives to Slot Size Admission Control

There are another two Admission Control policies. First is “Specify a Failover Host”, which dedicates a host (or hosts) for failover. Such host acts as a hot standby and can run VMs only in a failover situation. This policy is ideal if you want to reserve failover resources.

And the second is “Percentage of Cluster Resources”. Resources under this policy are reserved based on the percentage of total cluster resources. If you have five hosts in your cluster you can reserve 20% of resources (which is equal to one host) for failover.

This policy uses percentage of cluster resources, instead of slot sizes, and hence doesn’t have the issues of the “Host Failures Cluster Tolerates” policy. There is a gotcha, if you add another five hosts to your cluster, you will need to change reservation to 10%, which is often overlooked.

Conclusion

“Percentage of Cluster Resources” policy is recommended to use in most cases to avoid issues with slot sizes. What is important to understand is that the goal of this policy is just to guarantee that VMs with reservations can be restarted in a host failure scenario.

If a VM has no reservations, then “Percentage of Cluster Resources” policy will use only memory overhead of this VM in its calculations. Which is probably the most confusing part about Admission Control in general. But that’s a topic for the next blog post.

 

Dell Compellent iSCSI Configuration

November 20, 2015

I haven’t seen too many blog posts on how to configure Compellent for iSCSI. And there seem to be some confusion on what the best practices for iSCSI are. I hope I can shed some light on it by sharing my experience.

In this post I want to talk specifically about the Windows scenario, such as when you want to use it for Hyper-V. I used Windows Server 2012 R2, but the process is similar for other Windows Server versions.

Design Considerations

All iSCSI design considerations revolve around networking configuration. And two questions you need to ask yourself are, what your switch topology is going to look like and how you are going to configure your subnets. And it all typically boils down to two most common scenarios: two stacked switches and one subnet or two standalone switches and two subnets. I could not find a specific recommendation from Dell on whether it should be one or two subnets, so I assume that both scenarios are supported.

Worth mentioning that Compellent uses a concept of Fault Domains to group front-end ports that are connected to the same Ethernet network. Which means that you will have one fault domain in the one subnet scenario and two fault domains in the two subnets scenario.

For iSCSI target ports discovery from the hosts, you need to configure a Control Port on the Compellent. Control Port has its own IP address and one Control Port is configured per Fault Domain. When server targets iSCSI port IP address, it automatically discovers all ports in the fault domain. In other words, instead of using IPs configured on the Compellent iSCSI ports, you’ll need to use Control Port IP for iSCSI target discovery.

Compellent iSCSI Configuration

In my case I had two stacked switches, so I chose to use one iSCSI subnet. This translates into one Fault Domain and one Control Port on the Compellent.

IP settings for iSCSI ports can be configured at Storage Management > System > Setup > Configure iSCSI IO Cards.

iscsi_ports

To create and assign Fault Domains go to Storage Management > System > Setup > Configure Local Ports > Edit Fault Domains. From there select your fault domain and click Edit Fault Domain. On IP Settings tab you will find iSCSI Control Port IP address settings.

local_ports

control_port

Host MPIO Configuration

On the Windows Server start by installing Multipath I/O feature. Then go to MPIO Control Panel and add support for iSCSI devices. After a reboot you will see MSFT2005iSCSIBusType_0x9 in the list of supported devices. This step is important. If you don’t do that, then when you map a Compellent disk to the hosts, instead of one disk you will see multiple copies of the same disk device in Device Manager (one per path).

add_iscsi

iscsi_added

Host iSCSI Configuration

To connect hosts to the storage array, open iSCSI Initiator Properties and add your Control Port to iSCSI targets. On the list of discovered targets you should see four Compellent iSCSI ports.

Next step is to connect initiators to the targets. This is where it is easy to make a mistake. In my scenario I have one iSCSI subnet, which means that each of the two host NICs can talk to all four array iSCSI ports. As a result I should have 2 host ports x 4 array ports = 8 paths. To accomplish that, on the Targets tab I have to connect each initiator IP to each target port, by clicking Connect button twice for each target and selecting one initiator IP and then the other.

iscsi_targets

discovered_targets

connect_targets

Compellent Volume Mapping

Once all hosts are logged in to the array, go back to Storage Manager and add servers to the inventory by clicking on Servers > Create Server. You should see hosts iSCSI adapters in the list already. Make sure to assign correct host type. I chose Windows 2012 Hyper-V.

 

add_servers

It is also a best practice to create a Server Cluster container and add all hosts into it if you are deploying a Hyper-V or a vSphere cluster. This guarantees consistent LUN IDs across all hosts when LUN is mapped to a Server Cluster object.

From here you can create your volumes and map them to the Server Cluster.

Check iSCSI Paths

To make sure that multipathing is configured correctly, use “mpclaim” to show I/O paths. As you can see, even though we have 8 paths to the storage array, we can see only 4 paths to each LUN.

io_paths

Arrays such as EMC VNX and NetApp FAS use Asymmetric Logical Unit Access (ALUA), where LUN is owned by only one controller, but presented through both. Then paths to the owning controller are marked as Active/Optimized and paths to the non-owning controller are marked as Active/Non-Optimized and are used only if owning controller fails.

Compellent is different. Instead of ALUA it uses iSCSI Redirection to move traffic to a surviving controller in a failover situation and does not need to present the LUN through both controllers. This is why you see 4 paths instead of 8, which would be the case if we used an ALUA array.

References

Zerto Overview

March 6, 2014

zerto-logoZerto is a VM replication product which works on a hypervisor level. In contrast to array level replication, which SRM has been using for a long time, it eliminates storage array from the equation and all the complexities which used to come along with it (SRAs, splitting the LUNs for replicated and non-replicated VMs, potential incompatibilities between the orchestrated components, etc).

Basic Operation

Zerto consists of two components: ZVM (Zerto Virtual Manger) and VRA (Virtual Replication Appliance). VRAs are VMs that need to be installed on each ESXi host within the vCenter environment (performed in automated fashion from within ZVM console). ZVM manages VRAs and all the replication settings and is installed one per vCenter. VRA mirrors protected VMs I/O operations to the recovery site. VMs are grouped in VPGs (Virtual Protection Groups), which can be used as a consistency group or just a container.

Protected VMs can be preseeded  to DR site. But what Zerto essentially does is it replicates VM disks to any datastore on recovery site where you point it to and then tracks changes in what is called a journal volume. Journal is created for each VM and is kept as a VMDK within the “ZeRTO volumes” folder on a target datastore. Every few seconds Zerto creates checkpoints on a journal, which serve as crash consistent recovery points. So you can recover to any point in time, with a few seconds granularity. You can set the journal length in hours, depending on how far you potentially would want to go back. It can be anywhere between 1 and 120 hours.Data-Replication-over-WAN

VMs are kept unregistered from vCenter on DR site and VM configuration data is kept in Zerto repository. Which essentially means that if an outage happens and something goes really wrong and Zerto fails to bring up VMs on DR site you will need to recreate VMs manually. But since VMDKs themselves are kept in original format you will still be able to attach them to VMs and power them on.

Failover Scenarios

There are four failover scenarios within Zerto:

  • Move Operation – VMs are shut down on production site, unregistered from inventory, powered on at DR site and protection is reversed if you decide to do so. If you choose not to reverse protection, VMs are completely removed from production site and VPG is marked as “Needs Configuration”. This scenario can be seen as a planned migration of VMs between the sites and needs both sites to be healthy and operational.
  • Failover Operation – is used in disaster scenario when production site might be unavailable. In this case Zerto brings up protected VMs on DR site, but it does not try to remove VMs from production site inventory and leave them as is. If production site is still accessible you can optionally select to shutdown VMs. You cannot automatically reverse protection in this scenario, VPG is marked as “Needs Configuration” and can be activated later. And when it is activated, Zerto does all the clean up operations on the former production site: shuts down VMs (if they haven’t been already), unregister them from inventory and move to VRA folder on the datastore.
  • Failover Test Operation – this is for failover testing and brings up VMs on DR site in a configured bubble network which is normally not uplinked to any physical network. VMs continue to run on both sites. Note that VMs disk files in this scenario are not moved to VMs folders (as in two previous scenarios) and are just connected from VRA VM folder. You would also notice that Zerto created second journal volume which is called “scratch” journal. Changes to the VM that is running on DR site are saved to this journal while it’s being tested.
  • Clone Operation – VMs are cloned on DR site and connected to network. VMs are not automatically powered on to prevent potential network conflicts. This can be used for instance in DR site testing, when you want to check actual networking connectivity, instead of connecting VMs to an isolated network. Or for implementing backups, cloned environment for applications testing, etc.

Zerto Journal Sizing

By default journal history is configured as 4 hours and journal size is unlimited. Depending on data change rate within the VM journal can be smaller or larger. 15GB is approximately enough storage to support a virtual machine with 1TB of storage, assuming a 10% change rate per day with four hours of journal history saved. Zerto has a Journal Sizing Tool which helps to size journals. You can create a separate journal datastore as well.

Zerto compared to VMware Replication and SRM

There are several replication products in the market from VMware. Standalone VMware replication, VMware replication + SRM orchestraion and SRM array-based replication. If you want to know more on how they compare to Zerto, you can read the articles mentioned in references below. One apparent Zerto advantage, which I want to mention here, is integration with vCloud Director, which is essential for cloud providers who offer DRaaS solutions. SRM has no vCloud Director support.

References

Overview of NetApp Replication and HA features

August 9, 2013

NetApp has quite a bit of features related to replication and clustering:

  • HA pairs (including mirrored HA pairs)
  • Aggregate mirroring with SyncMirror
  • MetroCluster (Fabric and Stretched)
  • SnapMirror (Sync, Semi-Sync, Async)

It’s easy to get lost here. So lets try to understand what goes where.

Simple-Metrocluster

SnapMirror

SnapMirror is a volume level replication, which normally works over IP network (SnapMirror can work over FC but only with FC-VI cards and it is not widely used).

Asynchronous version of SnapMirror replicates data according to schedule. SnapMiror Sync uses NVLOGM shipping (described briefly in my previous post) to synchronously replicate data between two storage systems. SnapMirror Semi-Sync is in between and synchronizes writes on Consistency Point (CP) level.

SnapMirror provides protection from data corruption inside a volume. But with SnapMirror you don’t have automatic failover of any sort. You need to break SnapMirror relationship and present data to clients manually. Then resynchronize volumes when problem is fixed.

SyncMirror

SyncMirror mirror aggregates and work on a RAID level. You can configure mirroring between two shelves of the same system and prevent an outage in case of a shelf failure.

SyncMirror uses a concept of plexes to describe mirrored copies of data. You have two plexes: plex0 and plex1. Each plex consists of disks from a separate pool: pool0 or pool1. Disks are assigned to pools depending on cabling. Disks in each of the pools must be in separate shelves to ensure high availability. Once shelves are cabled, you enable SyncMiror and create a mirrored aggregate using the following syntax:

> aggr create aggr_name -m -d disk-list -d disk-list

HA Pair

HA Pair is basically two controllers which both have connection to their own and partner shelves. When one of the controllers fails, the other one takes over. It’s called Cluster Failover (CFO). Controller NVRAMs are mirrored over NVRAM interconnect link. So even the data which hasn’t been committed to disks isn’t lost.

MetroCluster

MetroCluster provides failover on a storage system level. It uses the same SyncMirror feature beneath it to mirror data between two storage systems (instead of two shelves of the same system as in pure SyncMirror implementation). Now even if a storage controller fails together with all of its storage, you are safe. The other system takes over and continues to service requests.

HA Pair can’t failover when disk shelf fails, because partner doesn’t have a copy to service requests from.

Mirrored HA Pair

You can think of a Mirrored HA Pair as HA Pair with SyncMirror between the systems. You can implement almost the same configuration on HA pair with SyncMirror inside (not between) the system. Because the odds of the whole storage system (controller + shelves) going down is highly unlike. But it can give you more peace of mind if it’s mirrored between two system.

It cannot failover like MetroCluster, when one of the storage systems goes down. The whole process is manual. The reasonable question here is why it cannot failover if it has a copy of all the data? Because MetroCluster is a separate functionality, which performs all the checks and carry out a cutover to a mirror. It’s called Cluster Failover on Disaster (CFOD). SyncMirror is only a mirroring facility and doesn’t even know that cluster exists.

Further Reading

Windows MPIO with IBM storage

September 17, 2012

IBM mid-range storage systems (like DS3950) work in active/passive mode. It means that access to each LUN is given through one controller, in constrast to active/active storage where data between host and two controllers can flow in round-robin fashion. So redundant path here is used only as a failover. Software which provides this failover functionality is called Multipath I/O (MPIO) and has implementations for all operating systems. I’ll desribe how to configure MPIO version for Windows.

Installation

Prior to Windows Server 2008, Microsoft didn’t have its own MPIO implementation and MPIO was distributed with IBM DS Storage Manager product. Now you can install MPIO from “Feautures” sub-menu of Windows Server 2008 Server Manager. After installation is complete you will find MPIO configuration options under Control Panel and in Administrative Tools.

IBM storage works well with default Windows MPIO implementation, however it’s recommended to install IBM MPIO (device-specific module) from Storage Manager installation bundle. In my case MPIO installation file was called SMIA-WSX64-01.03.0305.0608.

Enable multipathing

Initially you will see two hard drives for each LUN in Device Manager. You can enable MPIO for particular hardware ID (in other words, storage system) on Discover Multi-Paths tab of MPIO control panel. You can’t do that with LUN granularity. After you add selected devices and reboot, you will see them on “MPIO Devices” tab. Now each LUN will be seen as a single hard drive in Device Manager.

Configure preferred path

MPIO supports several load-balancing policies, which are configured on a LUN basis from MPIO tab of a hard drive in Device Manager. As a Load Balance Policy select Fail Over Only. Then for each path select which is Active/Optimized and which is a Standby path. Also make active path Preferred, so that after failover it failbacks to it.

Don’t be confused by iSCSI on the figure. It’s the same for pure FC. It’s just for reference.

Check configuration

When you configure active and passive paths you assume that first path listed is to controller A and second path is to controller B. But, in fact, there is no indication of that from the configuration page and you can neither confirm nor deny it. The only ID you see is adapter ports but they don’t even map to the actual ports on HBAs.

To be able to check your configuration you need to install IBM SMdevices utility which comes with IBM DS Storage Manager. Run DS SM installation and go for Custom Installation. There you need to check only the Utilities part. In SMdevices output you can see which path is preferred for this LUN and if it’s configured as active (In Use):

C:\Program Files\IBM_DS\util>SMdevices
IBM System Storage DS Storage Manager Devices
. . .
\\.\PHYSICALDRIVE1 [Storage Subsystem ITSO5300, Logical
Drive 1, LUN 0, Logical Drive ID
<600a0b80002904de000005a246d82e17>, Preferred Path
(Controller-A): In Use]

References

The best reference I found on that topic is IBM Midrange System Storage Hardware Guide (SG24-7676-01), from p.453: DS5000 logical drive representation in Windows Server 2008. As well as Installing and Configuring MPIO guide from Microsoft.