VMware ESX Server 3.0.2 is out. I applied it to my test environment (two Dell PowerEdge 1850s and a 650 as the VirtualCenter server) and it appears stable. It also didn’t wipe out my NTP configuration this time, which I like. I’ll run it for two weeks with artificial loads on it to make sure nothing is seriously wrong before I deploy it to production.
The release notes document some of the odd behaviour we’ve been seeing. First, Red Hat Enterprise Linux 5 has a problem where filesystems go read-only. This was also a problem with Red Hat Enterprise Linux 4 and is fixed now, but the fix didn’t end up in the RHEL 5 kernel tree. That KB article has a driver patch which helps.
We’ve also experienced HA shutting our VMs down when network outages occur, and they have a KB article on it. For us it’s led to disabling HA when we know network maintenance is happening. The settings for a cluster’s HA permits you to change the Isolation Response on a per-VM basis, too, but it always defaults to “Power off.” There are some forum threads on this (links below), which indicate that this will be addressed in the future with more options for heartbeats. It also suggests setting das.isolationaddress to an address that the HA node can ping successfully when it cannot reach the other nodes. I’m not sure what I’d set that to (127.0.0.1, ha!).
For me, a configurable timeout before HA kicks in would be nice, as well as a configurable default isolation response. If I could tell it to chill out for ten minutes before doing anything that’d be wonderful.
http://www.vmware.com/community/thread.jspa?threadID=87419
http://www.vmware.com/community/message.jspa?messageID=626646
http://www.vmware.com/community/thread.jspa?threadID=89335&tstart=0
You can select from a quicklist of the people you contact most, search your contacts, and manage multiple conversations. ,