Details
Release Date: 4/10/2008
Document Last Updated: 4/10/2008
Download Size: 206 MB Download Filename: ESX350-200803202-UG.zip md5sum: 62189e9967144cdaecdffba3035dfe66 |
|
Summary
This patch provides fixes for the following issues:
- Virtual machines that are powered off cannot be unregistered from an ESX Server host that is in maintenance mode.
- Support for virtual machine names with non-ASCII characters.
- When deleting snapshots, virtual machines with heavy I/O load might cause the hostd service to stop responding. This issue might also result in the VI Client getting disconnected from the ESX Server host.
- VMware VMotion completed successfully but memory resources on the source ESX Server host are not released.
- Failed SCSI release command can crash ESX Server host. The fix is for correctly handling the failure on release command.
- When performing a VMware VMotion to a VMvisor host in an over-committed-memory situation, the destination host stops responding and the failed migration does not recover on the source host.
- Fixes a network connectivity issue where packets with an 802.2LLC header (without SubNetwork Access Protocol (SNAP)) are dropped by the vmxnet driver in ESX Server 3.5. This issue has been observed with NetBEUI and SNA/DLC protocols.
- Virtual machine Power User cannot connect to a remote CD-ROM or floppy device.
- ESX Server host stops responding due to insufficient COW heap size. This fix allows users to configure the maximum COW heap size.
- Some OEM installation CDs won't work. The fix is for Microsoft Windows Virtual System Locked Pre-installation (SLP) certificate from the second to the first OEM SMBIOS string.
- Message Signaled Interrupts (MSI/MSI-X) are not fully supported on ESX Server 3.5. This fix disables MSI/MSI-X by default unless NetQueue is enabled.
- The Cisco Discovery Protocol (CDP) parsing code in the VMkernel does not support VLAN tags on CDP packets. As a result, all CDP packets received with VLAN tags are ignored and ESX Sever 3.5 does not detect CDP information. The CDP parsing code in the VMkernel did not support VLAN tags on CDP packets. The issue has been observed on some Cisco switches when the Native VLAN is changed to something other than 1.
In rare cases, virtual machines running certain guest operating systems, for example SUSE Linux Enterprise Server SP3, might be terminated.
- CDP is enabled (listen mode) on the vswitch by default when it is created. This results in passing of Virtual Infrastructure Client and VirtualCenter Server garbage values in the OID field of the CDP packets, which causes the VI Client and the VirtualCenter Server to stop responding.
- Allows Dell OpenManage Array Manager (OMAM) to report events from the RAID controller accurately.
- Stopping and starting the mgmt-vmware service might cause an unexpected reboot of virtual machines that are configured to automatically stop or start.
Solution
Note: Before you begin please refer to KB1003490 for important information on restarting the mgmt-vmware service.
Symptoms
- Trying to unregister virtual machines on ESX Server hosts without this fix, generates an error message:
A general system error: Invalid fault - No user-visible symptoms.
While deleting snapshots, the hostd stops responding. This error might be displayed:
Request failed because the remote server took too long to respond- The VMX process does not exit on the source ESX Server host. The message vmx| VMX has left the building: 0.might be missing from the end of the source VMX's log file.
- ESX Server host crashes.
- No user-visible symptoms.
- When running ESX Server in debug mode, each dropped packet produces the message
drop %p Net: Vmxnet2VMKDevPollTxRing:2523: drop 0x75fbcc0 - The issue with connectivity to a remote CD-ROM or floppy device might happen in this scenario:
- Assign virtual machine "Power User" role to a test user on a single virtual machine
- Log in to the VI Client as test user.
- Select the virtual machine, click on the Console tab, click the Connect CD/DVD button. The operation might fail with the error:
Vmomi.Fault.NoPermission: Permission to perform this operation was denied
- In a situation where the COW Heap gets exhausted quickly, warning messages are generated from COW code, indicating the COW heap is low, followed by a crash of the ESX Server host. The VMkernel warning messages might look similar to:
Mar 19 16:12:43 localhost vmkernel: 0:00:20:57.137 cpu2:1077)Heap: 1856: cow: heap below 4% -- 10968128 bytes free
Mar 19 16:12:43 localhost vmkernel: 0:00:20:57.088 cpu2:1086)WARNING: Heap: 1445: Heap cow could not be grown to accomodate the memory request
Mar 19 16:12:43 localhost vmkernel: 0:00:20:57.088 cpu2:1086)WARNING: Heap: 1522: Heap_Align(cow, 458752/458752 bytes, 4 align) failed. caller: 0x9da248
Mar 19 16:12:43 localhost vmkernel: 0:00:20:57.088 cpu2:1086)WARNING: Cow: 576: No memory available! Called from 0x9da248
When the COW heap is exhausted you won't be able to power on virtual machines. So when they cannot power on a virtual machine, you should check the var/log/vmkernel file to see if there are any messages indicating that COW heap has been exhausted. Based on this, the COW heap can be reconfigured. - Some OEM install CDs won't install.
- No user-visible symptoms.
- No user-visible symptoms.
- In rare cases the virtual machine running certain guest operating systems, for example, SUSE Linux Enterprise Server SP3, might terminate with this error in the vmware.log file:
vcpu-0:VMM fault: regs=0x2f8c, exc=13, eip=0x10da8b - No user-visible symptoms.
- The status of failed and rebuilding drives is not displayed correctly in Dell OMAM.
- Manually stopping or starting the mgmt-vmware service (hostd) through the service console causes virtual machines that are set to automatically start or stop to unexpectedly reboot.
Deployment Considerations
You must also install ESX350-200803208-UG to complete the fix for the issues addressed by this patch.
Patch Download and Installation
Based on VMware KB 1003696
