[ NOTE: machine translation with the help of DeepL translator without additional proofreading and spell checking ]
With the release of the new VM Analytics Appliance version 3.0, the article has been updated to summarize the changed system requirements and new features.
Major Feature Enhancements Version 3.0
The main innovation with VM Analytics in V3 is the extended in-guest/VM metrics. It is now possible not only to analyze capacities at datastore/vdisk level, but also to look into the virtual disks down to the guest operating system (partitions) with the help of VMware tools. In my eyes, this is another powerful free extension within the Collector and the Pure Storage Ecosystem.
In addition, it is now also possible with the appliance to upgrade the collector in-place in the future, as is usual with Pure (NDU), without having to redeploy.
Storage analyses or troubleshooting are not necessarily complex and time-consuming. With Pure1 and an appliance "VM Collector", analyses are possible beyond the storage layer deep into the VM/hypervisor layer. Identify bottlenecks in storage, SAN or hypervisor within minutes.
With VM Analytics, IO paths from VMs to FlashArray are visually represented. Means the entire path from: Disk (vDisk), VM, Host, Datastore, Storage Volume up to the Array. Pure Storage manages with "VM Analytics" consolidated storage and hypervisor monitoring* (disk monitoring only) without a 3rd-party monitoring tool. Within VM Topology, for example, you can filter by metrics such as IOPS, bandwidth and latency.
To run VM Analytics, a few requirements must be met:
the hypervisor for the appliance must be from VMware and at least a vCenter 6.5 U3 or newer. However, metrics can also be collected from vCenter 5.5 onwards
the VM Collector must be able to communicate with the vCenter (IP connection)
a service account with "read-only rights" (sufficient) is required
VMware Tools 10.1 or newer to query the in-guest VM metrics (capacities) (modified with UPDATE). - See also: FlashArray: VM Topology Requirements Checker PowerShell script
Phonehome must be enabled (port 443/HTTPS TCP outbound) to 188.8.131.52/27 and *.cloud-support.purestorage.com
It is possible to deploy the VM Collector in two variants:
Variant 1: OVA-Collector (this post)
A VMware OVA is deployed in the virtual environment. This requires manageable resources: 4vCPUs, 8GB RAM (changed with UPDATE) (with version 2 still 2vCPUs, 2GB RAM) and a vDisk with 40GB (also thin possible: after installation approx. 3GB are used directly). These resources are automatically allocated during the OVA import. In addition, experience has shown that it is advisable to place the VM Collector in a port group that has direct access to the management network of the flash arrays.
[NO LONGER] Variant 2: On-Array Collector in Purity Run
INFO: with the release of VM Collector as OVA the old version under the name of "Off-Array Collector" is no longer available for new installations.
There is also a migration option from Off-Array Collector to OVA and taking the existing configuration with you. For this, both collectors (old-new) must be able to reach each other via an IP connection.
One had the choice to run the VM Collector in the KVM stack of the controllers. Here, with the on-array variant, Purity had to be at the software level of at least 4.10.X, 5.1.0. The resources required are also minimal and do not affect system performance at all. Purity Run is already active by default since 5.2.X and therefore does not consume any additional resources. However, there are customers who do not want to keep an additional load on storage. Deployment of the On-Array Collector without the Pure Storage Support is not possible! For this a ticket would have to be opened and the employee would set this up for your system.
The Collector makes system queries to the vCenter every 10 minutes (600 seconds) by default, this interval is enough to query about 8000 VMs (modified with UPDATE). It is estimated that the load on the vCenter is less than 5% (on top CPU). The smallest possible value for the system queries is 5 minutes (300 seconds).
The VM Collector is based on Docker container architecture. It is possible to deploy a single collector on multiple vCenters within a network.
The latest OVA Collector can be downloaded from the Pure1 portal under Anaytics > VM Topology > "Gear" > "Download OVA Collector v3.0.0". If the vCenter has a prepared internet breakout and can access https pages, you can also just copy the download link and use it to deploy the OVF - if you know OVF templates, you know it's a next-next-finish setup with a few entries:
You can simply copy-paste the download link to the OVA into the setup (see requirements above).
In the further steps, a name for the VM, a datastore, EULA and network information must be assigned. The wizard is self-explanatory.
After that, you can start the deployment of the appliance by clicking "Finish".
After successfully deploying and booting the appliance, final configurations can be made via the VMware remote console or SSH. Finally, the default password must be changed and the vCenter instance must be specified.
The default user for the initial login is as usual:
The password must be changed at the initial login!
The "Authorization Key" can be created with "Create Collector" in Pure1. The name is just an alias, but as always I recommend to keep it unique. It is recommended to use a pattern like: "vCenterName-Collector-ArrayNames".
I have switched to a Putty SSH session in the meantime, since I can simply copy-paste the key here.
The registration of the connection Appliance-Pure1 is then done with the command:
purevmanalytics register Ihr-Authorisation-Key
HINT: Ich bin inzwischen auf eine Putty-SSH-Sitzung umgestiegen, da ich hier den Schlüssel einfach per Copy-Paste einfügen kann.
Die Anmeldung der Verbindung Appliance-Pure1 erfolgt dann mit dem Befehl:
After successful registration it is recommended to test the Pure1 connection via:
purevmanalytics test pinghome