Virtualization Adapted Adapting Business Processes for Virtual Infrastrcuture (and vice-versa)


VMware ESX High Availability – Tips and Tricks

VMware HA doesn’t work.

  1. Verify that host name is lowercase: hostname; hostname -s
  2. Verify that host name in /etc/hosts is lowercase
  3. Verify that search domain in /etc/resolv.conf is in lowercase
  4. Verify that host name in /etc/sysconfig/network is fqdn, all lowercase
  5. Verify that the host name in esx.conf is fqdn, all lowercase
  6. Verify that host name in DNS is lowercase: nslookup; <short hostname> (should properly resolve fqdn of host, all lowercase)
  7. Verify that all primary service consoles have the same name.
  8. Verify that all primary service consoles are in the same IP subnet.
  9. If VMotion vmkernel port is on same vSwitch as primary service console, use das.allowVmotionNetworks=1
  10. If host has multiple service consoles, use KB 1006541 and the das.allowNetwork0 HA option to ensure that only the primary service console is allowed.
  11. Verify that customer has appropriate licensing for HA, and has available licenses:  In LM Tools, perform a status inquiry, verify that cu is licensed for VC_DAS
  12. Once you have met all of the above criteria, enable HA.
  13. If, after you have verified all the above, and HA still won’t configure:
  1. On the host, stop vpxa: service vmware-vpxa stop
  2. The host will show not responding in VC after a while
  3. Disconnect the host from VC
  4. Re-connect the host to VC
  5. This will force the VPXA package to re-deploy, as well as the HA packages to re-deploy.
  6. Re-configure the hosts for HA.

Many thanks to: Kevin Riley []

See also:
As of VirtualCenter 2.5 Update 2 configuration of VMware High Availability fails.
An error similar to the following appears in the Tasks and Events detail:

HA agent on <esxhostname> in cluster <clustername> in <datacenter> has an error Incompatible HA Networks:

Cluster has network(s) missing on host: x.x.x.x

Consider using the Advanced Cluster Settings das.allowNetwork to control network usage.
– Allows for a NIC that is used for VMotion networks to be considered
for VMware HA usage. This parameter enables a host that has only one
NIC configured for management and VMotion combined to be used in VMware
High Availability communication. By default, any VMotion network is
das.allowNetwork[…] – Allows the use of port group names
to control the networks used for VMware HA. The value is set as the
name of the portgroup, for example, Service Console or Management
Network . When configured, the VMware HA cluster only uses the
specified networks for VMware HA communication.

To configure VMware HA to use the new settings:
Log in to VirtualCenter with the VI Client as an administrator.
Edit the settings of the cluster and deselect Enable VMware HA.
Click OK, and wait for the servers to unconfigure for VMware HA.
ESX Server > Configuration > Networking on each of the ESX hosts
in the cluster and note the portgroups that are common between the
Edit the settings of the cluster, and select Enable VMware HA.
Click VMware HA.
Click Advanced Options.
Add the das.allowNetworkX option with a value of the portgroup name, where X is a number between 1 and 10,

IR: Wednesday, June 24, 2009


Server Room Infrastructure Information

Filed under: virtualization — Tags: , , , , , , , — iben @ 15:03

Take an average office turned into a server room.  Many companies throw a 3 or 5 ton split air conditioner on a wall call the electrician and start loading the racks up with servers. Now a days the new server hardware can use all the power (and cooling) you have in probably one rack (think three loaded blade chassis with 16 servers each = 90 amps of 208v).  The question is how much cooling can you afford?

For customers deploying 10 or more servers it is recommended to use the most efficient power delivery and structured wiring options.

Calculate you power and cooling needs with this spreadsheet.

Using 3 phase power reduces the number of PDUs and “whips” required to power up large quantity of machines.  The actual cables from the server to the PDU are the same and the power supplies still receive 1 phase 208v power.  This is just a different and more efficient method of delivering the power to the servers.

You get higher densities with fewer wires to the electrical service panel with 3 phase power over single phase power. ServerTech has a great white paper on this topic here:

A 3 phase 30 amp 208V circuit can deliver 8.6 kw versus a 1 phase 30 amp 208V circuit which only delivers 4.99 kv.

Here is one of the more popular units…

Go with the 4 wire “Delta” configuration instead of the 5 wire “WYE” config. Wye power can distribute both 208 V and 120 V power from the same cabinet power distribution unit but requires an extra wire per PDU and is not needed in most datacenters. You will also be limiting the number of 208 volt power outlets with a WYE config PDU.

CS-24VD-L1530 – List $860.
CS-48VDD-L1530 – List $985.
CS-84VDD-L1530 – List $1,290.

You need to hire an electrician at $100 per hour x 4 hours plus parts.

Cheap fast switch with life time warranty:

24 ports NetGear GigE Unmanaged Switch – $260

SuperGoose temp and humidity – $500

Trendpoint power monitoring – direct $3085

48 port tie lines from each server rack to a central control rack cost about $2000 per rack

Sample power calculations (need to be adjusted) 24 amps x 208 volts = 5000 watts * 3 racks = 15000 watts = 51,113.088 with BTU/hour = 4.25 tons cooling (worst case)

3 Homaco M6 Square Hole 4 post Racks – $360.19 each
OR-19-84-SSDA2732. 84″- M6 SERVER RACK. 84″- M6 SERVER RACK.

And you can get all the various power cables needed for 208vac
operation from quail.

Powered by WordPress