Today has been another busy day, it was a tiring one yesterday and I was definitely down for more of the same today. No rest for the wicked…
I had a bad start, missing the session I wanted to go to for AppVolumes with Andrew Johnson. This was my fault as I hadn’t given myself enough time before the session to work out that it was actually located in the Marriott hotel nearby. It’s a huge hotel with many floors and the signs at the lobby didn’t help me find where I needed to go. It wasn’t the end of the world however, as I found a session close by for vSphere Storage. I’ll post below some of the key points that I took away from this session.
The rest of my day has been quite exciting. I managed to meet a few vHero’s of mine today. I had my Essential VSAN book signed by the legends that are Cormac Hogan and Duncan Epping. Really nice guys, I wish I stood and talked to them more but I didn’t want to take too much of their time and was a little star-struck, embarrassingly. Only after walking away I realised I could have had some conversations about VSAN, I even have some quesions for them!!! Oh well, I might see if I can catch them later. Here is a picture of us at the book signing:
Moving on to meeting more excellent guys, I was walking out of the lunch hall and looked to my right and saw Chris Wahl! As I was on a roll I thought I’d say hello and thank him for his contributions to the community. He was a thoroughly nice chap, I’m looking forward to perhaps bumping into him again.
In the spirit of VMworld, as I was writing this very blog post in the VMworld Bloggers section, Mr. Joep Piscaer (@jpiscaer) sits down next to me. As expected, he’s a top guy and we had a great chat about Nutanix, VCDX and the nuances of Hyper-VCDX (ask him!). I have thanked him many times in my Nutanix blog posts, it was great to meet him in person!
Finally to end the day before parties, I went to the vCSA Best Practices presented by William Lam. I’ve blogged a bit about the vCS to vCSA in my home lab and am interested in converting other environments. This session did not disappoint and I came away with some extra info and also the tech preview of the new migration appliance #migrate2vcsa.
I took notes on both the sessions I attended today and they can be found below.
vSphere Storage: Meet the Experts (VMFS, NFS, PSA, iSCSI) – Christos Karamanolis
ESXi 6 has improvements in the storage stack.
– Multiple completion worlds in PSA
– Deferred pool execution in PSA
(It is off by default expert mode advanced config can be enabled.
– Increases parallelism in parts of the storage stacks (Issuing compete I.O commands)
– Most beneficial interfaces with large cost per issuing IO (AHCI)
Performance tweaks for PCIe SSD
– Use multiple PVSCSI adapters – the more controllers you have the more aggregated IO you can have down to the hypervisor.
– Increases PVSCSI queue depth to 255
– Increase maximum disk requests to 255
– Use higher end configs with more than 6 cores per socket.
– Enable SCSI Deferred pool via ESXCLI – off by default as above.
– IPv6 support
– Software, Independent and dependant iSCSI and also booting from IPv6 target
– Independent HBA support DHCP through UI and ESXCLI
– Software iSCSI now has static route support. Routing with port binding is supported.
Support starting vSphere 6
– Mount an NFS volume as NFS v4 or v4.1 NOT BOTH
– Supports multi-pathing for servers supporting session trunking
– Supports Kerberos authentication protocol to secure communication
– Supports non root users to access files when used with Kerberos
– Supports iPV6 with AUTH_SYS, not with Kerberos yet
– Upgrade path – using svMotion or unmount from NFSv3 and mount as v4.1 volume.
– Shared disk clustered file system optimized for MVs
– Enable advanced capability’s like snapshots , thin provisioning, DRS/SDRS.
Note: VMFS-3 is DEPRECATED starting vSphere 6.0 can no longer create them and upgrade is strongly recommended.
512e/4kn disk support
– Metadata aligned and sized for 4k sector disks
– Avoid RW on 512e disks for metadata I/O
Fast creation of large thick files
– Swap files and thick files
– Allow faster boot up of monster VMs
– Minimize I/O when accessing large thick files
Support for volume larger than 64Tb
Support for file larger than 64Tb
VMware vCenter Server Appliance – Best Practices & Tips/Tricks by William Lam
Power on VM’s : 10,000
Hosts per Cluster: 64
VMs per cluster: 8000
Linked mode: Yes
These configuration maximums are true for both vCS and vCSA.
– In vCener 6.0 U1 the FULL VUM capabilities are in the vSphere web client.
– There 3 Deployment types: Embedded (PSC and VC) or standalone for either PSC or VC
When considering which deployment type to choose, it is worth knowing that the external PSC enables enhanced linked mode. This is a key consideration. If not then maybe keep things simple with embedded.
You can have a windows external PSC and a vCSA PSC replicate between each other seamlessly.
A common question is “Can I change from Embedded PSC to external PSC? The answer is in Update 1 , it will be possible to migrate from embedded to External PSC! The steps, simply put, are:
1) Deploy an external PSC
2) Replicate from embedded vCSA.
3) Point the VC component on embedded to the external PSC.
4) Use a reconfigure option using cmsso-util via CLI on the vCSA.
Note: This cannot be used to merge between two different SSO Domains.
In update 1 you can now deploy a vCSA to a vCenter or ESXi host, supporting brownfield and greenfield deployment types.
When deploying it is important that proper DNS is configured and reachable from Client desktop and vCSA and all IPs are resolvable for forward and reverse lookups.
Ensure proper NTP time synchronization. Check time on ESXi host, PSC and Client desktop are all sync!
If you are using a VDS with your vCSA, the portgruop it uses must be ephemeral when using ESXi as a deployment target.
I posted about full releases and upgrading from 5.5 to 6.0 requiring a migration. BLOGGED.
For smaller Update releases (U1, U2, etc) there is now in-place upgrades available, just mount the ISO to the vCSA and it will upgrade with a brief service restart, causing minimal downtime.
Services can now be managed from the Web UI with useful Options to stop/start/restart services on the vCSA.
The VAMI (Admin Interface) has retuned in 6.0 Update 1 with a brand new HTML web interface which still runs on port 5480.
There is a brand new PSC interface UI in U1 that also uses HTML5.
Everything that can be done from the VAMI interface can be performed from the appliance shell (applancesh).
The appliancesh is now automation friendly by being able to accept multiple commands via CLI.
It is possible to change the deployment size with extra memory and cpu. Dynamic memory resize support with a reboot.
Easy to resize disk capacity now by increasing the VMFK size. Step 2 run vpxd_servicecfg command line to expand the disk.
All configuration parameters about how the appliance is deployed in /etc/vmware/install-defaults
VIMTOP can be used on the vCSA to view services, similar to TOP.
Support bundles available via SSH, or open a browser to be able to download: https://VC-IP/appliance/support-bundle
There is a fling for vCS to vCSA converter which I also posted about.
Supports 5.5 to 5.5 at present and supports MS SQL Server 2008 R2.
There is also a Tech Preview of an improved appliance to perform the conversion which will hopefully be coming soon!