Happy thanksgiving to my American colleagues!
after a bunch of “high level” of posts, glad to do a technical one again :-)
Glad to see vSphere 4 udpate 1 hit today – you can get more here.
A couple important things are improved/added – I’ll call out the key ones for me:
- Windows 7 and Windows 2008 R2 support (32-bit and 64-bit)
- MSCS gets VM HA and DRS support in a limited fashion – you can have MSCS nodes supported (worked before) in a cluster by excluding them from VM HA and DRS. This means no more “MSCS only VMware clusters”.
- paravirtualized SCSI support for boot disks on W2K3 and W2K8 (nice, eliminates one of the pvSCSI beefs)
- View 4 support (you need update 1 to use View 4)
- This update didn’t trigger re-cert on the HCL, so the stuff that was on the vSphere 4 HCL is there for update 1
A smaller item, but there’s an important fix in update 1 for folks using anything that updated certain things (one of which was any 3rd party Pluggable Storage Architecture SATP, PSP or MPP). These they occasionally caused the storage view tab to show the following error (Mike Laverick – you helped us fix this :-) :
The storage service is not initialized. Please try again later.
The following entries might be written to sms.log during the periodic SMS initialization cycle:
2009-04-20 12:18:57,180 [Thread-5] DEBUG com.vmware.vim.sms.provider.VcProviderImpl - Populating scsi volume information...
2009-04-20 12:18:57,289 [Thread-5] ERROR com.vmware.vim.sms.provider.VcProviderImpl - Failed populating service cache
com.ibatis.common.jdbc.exception.NestedSQLException:
Other important items:
- FCoE/10GbE
- NMP RR (changing the IOoperationlimit value) – important note!
- we’re getting reports that when customers change the IO Operation Limit from 1000 (default) to 1, on ESX reboot, weirdness happens. This parameter governs how many IO operations are sent before a new path is selected when using NMP Round Robin. On reboot if it has been changed from the default to a value of 1 – it becomes what seems a random value – making loadbalancing… less than optimal. I don’t recommend changing the value until this gets resolved – just leave it at the default. I expect that this should affect all arrays using block devices equally. It doesn’t affect PowerPath/VE users– as it is a 3rd party MPP – it replaces the full NMP stack (both the SATP and PSP – and this looks like a PSP thing). Will do a more detailed post on this early next week.
- VMware Update Manager (VUM) expanded in update 1 with more PowerPath/VE goodness.
- Want to see the power of the customer? I did a survey on Virtualgeek asking “how should we patch 3rd party vmkernel modules?”. Results? below:
Alright customers – you have spoken! PowerPath/VE 5.4 can be installed and even patched to PowerPath/VE 5.4.1 (will be released soon, and will of course be the model for subsequent releases) integrated with VUM. In vSphere 4 udpate 1, VUM gets a “extension” option for 3rd party baselines for vmkernel-level modules and patches (think PowerPath/VE, Cisco Nexus 1000v). See a demo below.
You can download the high-resolution demonstrations in WMV format and MOV format.
FYI – I know that people ideally want PowerPath/VE licensing to be integrated with the new vCenter-centric licensing model – working on it….
We just updated out vmware cluster to 4.0 U1 today, and we have been using those FCoE Qlogic drivers for a few weeks now and are very solid.
Posted by: David Robertson | November 20, 2009 at 08:01 PM
re: your last comment around licensing - this is key - licensing of all these type of products has been problematic for some time - and you need to make it easy for the end-customer (IT geeks typically don't like explaining complicated licensing models for software to the guys that hold the purse strings) - make it easy ;-)
Twitter me @stuiesav or use email me if ya want - cheers, Stuart.
Posted by: Stuart Savill | November 20, 2009 at 08:04 PM
Could the 3rd party VUM mechanism be used for things like the HP tools? The difficulty of installing and patching those is probably my biggest complaint with vSphere to this point.
Posted by: Andrew Fidel | December 03, 2009 at 02:15 AM