5 Essential Tips for Maximizing Your Experience at Nutanix .NEXT for Bloggers
@mikkisse is correct.Going to the VM and looking at the I/O metrics will give you a lot of information pertaining to Read/Write distribution on disk, as well as the type of writes. If doesn’t give long term I/O metrics, but if you’re seeing alerts, you can watch this and get an idea of what’s going on. Few recommendations as we run numerous MS-SQL servers on AHV and ESX clusters on hybrid clusters.Take advantage of AHV Turbo. This helps greatly when you’re on a hybrid cluster especially. When you have busy databases that require high I/O, make sure they exist on separate virtual disks. Doing so on AHV will allow for better CPU consumption and more efficient storage operations. The Wait is Over: AHV Turbo is Here! | Nutanix Community Make sure you have good indexing jobs. Not having proper indexes can hurt application performance.SQL Server Index Architecture and Design Guide - SQL Server | Microsoft Docs Make sure that your disks have the correct Allocations. Most SQL servers
We have for several hundred machines, what would you like to know? Installed on 2012 R2, 2016 and 2019 servers.
We've had to do this a couple of times. Great info. You can also achieve this through the Rest API. https://portal.nutanix.com/#/page/docs/details?targetId=API-Ref-AOS-v510:pro-api-abortReplication-auto-r.html
@arjun_sen this is most welcome news! Appreciate the article shared and very happy to see failover clustering supported!
Unfortunately the unmap and trim commands that esx issues doesn’t work on NFS based storage. you have to run a tool like sdelete or fill the drive and purge it to reclaim the space.
Technically you can defrag, and do so at the VM level. https://www.vmware.com/support/ws55/doc/ws_disk_defrag.html You really shouldn't defragment your virtual disks on Nutanix. It's bad for the SSD drives and the way storage extents work on Nutanix, it's unnecessary since it is already indexing and organizing data for you in 64k chunks. Your vdisk already has extents across multiple drives, so this will generate unnecessary IO. Of course I always recommend reaching out to support. They can provide you best practices.
Great write up!
Excellent. Thanks for sharing the video, hopefully this helps others with the situation you were in. Don't forget to mark your solution, so the post is marked as resolved.
Glad you found a solution! Which one worked for you if you don't mind sharing?
I've seen this on server 2016, where the image I downloaded was incorrect. You'll need to make sure you're downloading just the Server 2016 standard ISO or the datacenter ISO. If you're using an old OS, or have verified that your ISO's are correct, you may want to try the following: The link below will get you an ISO. When your building your first image, have two CD Drives, one for the OS image, another for the storage drivers. Where you see no disks displayed, load the second CD-ROM and inject the storage drivers. You should see your disks. http://download.nutanix.com/mobility/1.1.4/Nutanix-VirtIO-1.1.4.iso
Thats pretty much it. You'll want to read through VUM upgrade proceedures and get it configured properly, but at a high level that's the steps. Support will likely be more than happy to assist with the upgrade.
One click upgrades require DRS so that machines can vmotion from host to host and isolate the host receiving the upgrade. If you do not have DRS capabilities, you can use VUM to upgrade a host, or upgrade it manually, you will not be able to use one-click upgrades. Keeping in mind, that you will need to vacate the host of vms, and shut down the CVM on that host and then place the host in maintenance mode. I would always recommend filing a feature request with Nutanix Support for the ability to leverage one-click upgrades with your license type. We have done the same for our 36 ROBO clusters that have licenses that do not include DRS. We have to upgrade 100+ nodes manually and we spend a week plus doing upgrades for hypervisor, firmware, etc... because of this limitation. If you can sustain the outage, you can shutdown your vms and place all nodes into maintenance mode and kick off the upgrade from VUM of your vcenter is on a different cluster and can still reach the hosts. Th
If there is prolonged usage and you're using a TLC or even QLC SSD, I could see valid concern. The supermicro appliances typically use either SLC or MLC SSD drives. If a system is using SLC SSD's I'd be less concerned given how they are build for intense Read/Write operations beyond the consumer grade SSD. If a system is using TLC or QLC type SSD's, I would avoid using swap on them, as the storage density method, while much larger, degrades the cells degrades and the overall performance faster with each write and erase operation. I would overall agree with your assessment of avoiding using swap space on SSD as a general practice, with the caveat of knowing the type of drive being used.
A Linux best practice is to have your swap on a seperate vdisk from your root of "/" directory. Use LVM and break up your logical volumes as you wish, but I always make sure my swap is on a seperate vdisk from my root. It's meant as a last resort and you don't want the churn from swap taking away from OS performance. The same is true with Windows and your pagefile. The Microsoft best practice is to move your pagefile to a dedicated vdisk. Realistically in terms of ssd and hdd, we've never pinned our swap drives to ssd. You will see decreased performance on hdd, but it's meant as a temporary safety net, not a permanent solution for additional memory. If you did use SSD pinning for swap, I don't see how it would decrease ssd life unless you're constantly living in swap space. You're okay either way. Make sure you've enabled AHV Turbo (multi-queueing). For even better performance. I'll share a link once I'm at a desktop. Hope this helps!
LCM is definitely the route Nutanix is headed. I think they're wanting to move everything under that umbrella and make it centralized. We've had some issues in the past with BIOS and BMC updates, but we caught it early in the lab. After the spectre meltdown updates came out, the phoenix software didn't account for a second reboot applied after the update and it hosed the upgrade. So we had the manually finish the cluster. They've come a long way and as a consumer that's only good news! Glad things went well for you!
As [user=30892]nishantdhingra1[/user] reiterated with LCM, this will evacuate a host for you by obeying DRS. If you do not have DRS or running AHV, you must upgrade a node manually. In that case, yes you would need to evacuate your host before beginning the SATADOM upgrade. If you're put in that position, reference this guide. [url=https://portal.nutanix.com/#/page/docs/details?targetId=Hardware-Admin-Ref-AOS-v510:bre-manual-firmware-update-c.html][u]https://portal.nutanix.com/#/page/docs/details?targetId=Hardware-Admin-Ref-AOS-v510:bre-manual-firmware-update-c.html[/u][/url] If anyone has answered your question, please remember to mark their response as an answer, so that others can be assisted by your question in the future.
As a general rule of thumb, you never want to run over your HA limit on RAM and especially storage. You cannot over commit RAM anyways on AHV (For Now), but you can on ESXi. Basically you obey the N+1 mindset. If you have node awareness, you'll never want to provision more than your HA can handle of losing a single node. If block awareness, you'll never want to over provision over a full blocks worth of resources on memory and especially storage and so on and so forth. On any storage platform, once you exceed 80% storage utilization, it generally starts impacting IO capabilities. Some platforms "swear" you can go up to 90%, but you should make it a goal to never reach that high. I would spin it as "To ensure we maintain optimal performance, as well as ensuring we have resources available in the event of hardware failure, this is why we can't provision to complete capacity." He's likely getting pressure to use the resources paid for, but I would document out why what he's want
[user=9121]Jon[/user] I saw you liked the response in the mentioned article. [url=https://next.nutanix.com/server-virtualization-27/ahv-vcpu-vs-vcores-18216]https://next.nutanix.com/server-virtualization-27/ahv-vcpu-vs-vcores-18216[/url] I've always understood that you should never allocate vCPU in excess of your physical socket count. I'm genuinely curious how AHV handles that, and is it not really a concern from a CPU scheduling perspective?
VMWare, AHV, KVM, Hyper-V, it doesn't matter, you should never assign more sockets than you have available on a given host. This would introduce a CPU scheduling issue and you'll likely see CPU ready times increase as a result. If there isn't heavy load on the node you probably wouldn't notice, but you should use 1 socket and 4 cores if you have a single socket node. If you had 2 physical sockets on the host you could do 2 sockets with two cores each, but still not 4 sockets with 1 core each. Hope that makes sense.
Gotcha, that makes more sense now. It doesn't look like there is any sort of manual deployment option. I referenced the general guidelines here: https://portal.nutanix.com/#/page/docs/details?targetId=Web-Console-Guide-Prism-v510:wc-cloud-connect-wc-c.html#nconcept_lhz_jmf_ws I would definitely still engage support.
I am unaware of any product with direct access into the Nutanix API for what you're looking for. I would recommend looking at products that have a pre-install script capability such as Shavlik (Ivanti). https://help.ivanti.com/sh/help/en_US/PAT/23/Topics/Pre-Install_Script_and_Post-Install_Script_Tabs.htm You could have it call a script that reaches into the Nutanix REST API using a post command. [url=https://dc1-nutanix:9440/api/nutanix/v2/api_explorer/index.html#!/snapshots/createSnapshot][u]https://clustername:9440/api/nutanix/v2/api_explorer/index.html#!/snapshots/createSnapshot[/u][/url] You can reference the API call from here. [url=https://dc1-nutanix:9440/api/nutanix/v2/api_explorer/index.html#!/snapshots/createSnapshot][u]https://clustername:9440/api/nutanix/v2/api_explorer/index.html#!/snapshots/createSnapshot[/u][/url] Personally if I did this, I would want to make sure that upon completion of the snapshot, that you have the uuid of that particular sna
So my first question is, why would you want CVMs in Azure? Controller VMs are meant for storage I/O operations for your hardware and the Acropolis code. To my knowledge AWS is the only offering at this point to deploy Nutanix AOS for clustering capabilities for bare metal deployments. Essentially you are using EC2 resources to have an actual cluster on Amazon hardware. There are no such offerings in Azure. Cloud connect is intended to be a replication point for disaster recovery purposes. You're not replicating to another Nutanix cluster. I would highly recommend engaging support with your inquiry.
I believe the short answer is no. The only way to prevent a system from depending on flash tier storage is by making sure you're not pinning it. By design any persistent writes and reads will be placed on flash regardless of the configuration. This is just how the flow of data works. An extent written will be moved off of flash tier storage after a period of time when it's considered cold data. With reads, if a file is constantly being read it will be moved onto SSD to reduce read latency. You can pose the question to support, but I believe there is no command to bypass flash altogether.
Sometimes with documentation you'll find remnants that are outdated. Unless an application such as Microsoft Exchange explicitly call out for storage to not have dedupe turned or or compression, it's always a best practice. I would raise this up to support for a clear and official stance, but my personal opinion would be to turn on dedupe and compression. The caveat is understanding that you will need to increase RAM on your CVMs of you enable it. Compression on a highly active container can incur higher CPU overhead, so setting post-process compression is recommended. I have some clusters where we set it to compress almost a day after the initial write so I don't have the CPU overhead as high on my CVMs. Hope that helps.
Realistically, when we increased our core count it helped performance a lot. If you've already done so, I would say that your issue isn't CPU based. As far as testing stress on a SQL server, there are several internet resources that can help you simulate load on a SQL server. i.e. https://www.brentozar.com/archive/2015/05/how-to-fake-load-tests-with-sqlquerystress/ To simulate high Storage I/O you can use a test tool call IOMeter. https://www.itprotoday.com/mobile-management-and-security/using-iometer-measure-disk-performance I'm not sure what you're using to measure your current query times, but I assume it's SQL Management studio. Have the queries changed since moving it over? Above all else, if this is POC hardware, I would directly engage with your sales engineer and or support to understand what might be going on. There may be some optimizations that they recommend specific to your situation.
Already have an account? Login
Enter your username or e-mail address. We'll send you an e-mail with instructions to reset your password.
Sorry, we're still checking this file's contents to make sure it's safe to download. Please try again in a few minutes.
Sorry, our virus scanner detected that this file isn't safe to download.
