HA Insufficient resources to satisfy HA failover

[nigoool]

I was wondering if anyone else has seen this issue.

I have setup 4 ESX 3.0 nodes in a cluster with HA and DRS enabled. When you go to the Summary page for the Cluster on the VC it indicates that all is well and that the "Current Failover Capacity" is 4 and the "Configured Failover Capacity" is 1. This behaves itself for about 1 hour and then the value for the Current Failover Capacity changes to 0 and so the "HA Insufficient resources to satisfy HA failover" alert appears even though there is 4 hosts in the cluster and working fine.

Cheers Nigel

[mreferre]

I have seen something similar and my problem was that the host name (host+fqdn) was longer than 30 characters .... and this is a known bug of VMware HA. (KB 2259).

Massimo.

[nigoool]

Yeah I checked this KB article out but my FQDN are only 27 long...mmmm

It configures okay and the HA failover does work but for some reason after a period of time it just thinks there are 0 hosts for failover capacity. The only way to fix it is restart the VC service.

[mstahl75]

I just implemented HA last week and at first had the issue where the FQDN was too long. I used the steps outlined in the KB article and HA was showing as correctly configured. I thought I would check it today and it is now showing that it has insufficient resources. Restarting the VC service doesn't help for me.

It looks like there may be other issues in play here.

[mstahl75]

You might look at the files in /opt/LGTOaam512/log[/b] to help troubleshoot. I'm having the problem where fails after a time and from looking at some of the logs it looks like it is still having a problem with the hostname even though I have followed the KB recommendation to get it to work.

[dsolerdelcampo]

I have got the same message. I did not know where the problem was and my host name was 16 character long.

What I did was to display the pop up menu in my host and I selected the option "Reconfigure for HA". It did not work but it gave me a clue: I got the error message "Failed to communicate with the remote host, either due to network errors or because the host is not responding". I had installed this host upgrading my ESX 2.5.3 server, not installing form scratch and my service console port had been defined in the Virtual Switch 1, instead of in the vSwitch0 and besides showed that it was a legacy eth0.

What I did was define a new Service console port in the vSwith0, selected again the option "Reconfigure for HA" and it started to work properly.

Regards,

[MarkBK]

I have also have the status "Insufficent resources" for my HA cluster. We did have some DNS issues to begin with, but after resolving them the agents appear to be running correctly, but the cluster still says not enough resources.

We have 57Ghz of CPU, 95GB memory, 6 hosts, 24 processors, and 48 guests. Average CPU utilization is under 10%. Memory is running at 43%.

HA actually works, we had a ESX crash one night and everything restarted on another host, but it is annoying that VC says we don't have enough resources. I would like to think that the guests could all run on 4 hosts in a pinch.

My Fully Qualified hostnames are 21 characters.

HA should log somewhere why it thinks we are short.

Any suggestions?

[mstahl75]

HA should log somewhere why it thinks we are short.

I posted above where some of the logs for HA are stored. You might give a look in that directory from the COS. That's where I found out for sure that I was still having issues with the length of my FQDN.

[MarkBK]

I guess I did not think that those logs would help me understand why the cluster thinks there are not enough resources.

However, perhaps the agents are still not configured right. I notice in the logs that I have 5 primary agents and one secondary. I suppose that I should have one primary and 5 secondaries.

I am not sure if that is correct or what to do about it.

[mstahl75]

I'm not really sure. I had a log file called servername_agent.err which helped out. That was the one that was most obvious.

[MarkBK]

I am going to mark this thread as answered and pursue this in the other thread.

http://www.vmware.com/community/thread.jspa?threadID=48837

[MarkBK]

Whoops, this is not mine to close--sorry.

[boatrke1]

Noticed the same issue.

Current Failover Capacity: 2 hosts

Configured Failover Capacity: 1 hosts

After a virtual machine has been powered off, I receive the following:

Configuration Issues

Insufficient resources to satisfy HA failover level on cluster Cluster1 in ISDC.

Current Failover Capacity: 0 hosts

Configured Failover Capacity: 1 hosts

Restarting the vpxd service on the Virtual Center Server clears the error.

[aleneves]

Hi,

I correct the configuration of /etc/resolv.conf in host to resolv domain names and this error disapear... I hope that is help \!!!

[markzz]

Nigoool being the the thread initator was working with me on this.

Since that time we have upgraded the VC to Ver2.01 and the ESX hosts to 3.01 (except 1 host is still running 3.0 which has proven to exibit some very odd behaviour).

I'm not entirely sure our HA is configured correctly but since upgrading to VC Ver 2.01 it's no longer in error, it has however always worked. We too have had a number of Host failures. (CPU Lockups on all occasions but only on esx version 3.0, none on 3.01)

Although this isn't the approariate thread, I will make mention of the incompatabilities we have seen with VC 2.01. It works, but there are some odd behavioural issues, these occur when the guest is on the ver 3.0 host.

eg. Create a new Guest and the alocated disks will always be 1GB even though you sepcify other sizes.

Odd guest shutdown issues, I've had to kill a couple of guests as they didn't shutdown but the vmtools had stopped therefore I couldn't power then off.

A grey screen type hang on Windows 2k3. In this case as above the guests were not ping able.

I guess the point here is don't run a mixed version environment for longer than you must.

[markzz]

Another note.

Hi Nigel.. Merry Christmas

and a Merry Christmas to all.

Have a lovely time off. Turn that PC off, Get out of your chair, have a beer, go for a walk..

Hope Santa is kind to you all,

[jaspain]

I've been working with VMware TS on this issue, and believe it is solved. Almost all of our virtual machines were converted from ESX 2.x to ESX 3.x. Under ESX 2.x the default memory reservation was 1/2 the configured amount of RAM and the memory limit was equal to the configured amount of RAM. This appears to cause problems for HA in ESX 3.x. In ESX 3.x the memory reservation is zero, and the limit is "unlimited." To see this, edit the settings of a virtual machine, click on the Resources tab, and select Memory on the left. To conform to the ESX 3.x defaults, change the settings to a reservation of 0, and check the Unlimited box under limit. After doing this for all virtual machines, edit the settings for the cluster. Disable HA, and then edit the cluser settings again to reenable HA. The current failover capacity should now match the configured capacity.

[sbeaver]

This worked like a champ!! Good find and thanks for posting

[davidbarclay]

Good work, you just saved me a bunch of time (I will be doing this next week).

Is it just me or are there a lot of "issues" post-2.5 upgrade?

Dave

[jaspain]

I wouldn't say there were a lot of issues, but there was one other serious one that nailed us. The in-place conversion from VMFS2 to VMFS3 doesn't work on volumes greater than 1 TB in size. This is because VMware decreased the maximum block size in VMFS3 from 16 MB to 8 MB. We have a SAN that had a 2-TB VMFS2 volume with 16-MB blocks that were too large to convert. We ended up having to expand the SAN, create a new VMFS3 volume, and copy our virtual disk files. This cost real money and took many hours...