• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

VMware HA agent connection issues

T

Thuleman

Supreme [H]ardness
Joined
Apr 13, 2004
Messages
5,833
I don't want to put in an SR for that just yet because I don't do "best practices" running my vCenter in a VM, but I didn't have this problem in 3.5

In vSphere my cluster doesn't work right. All three boxes are sitting on the same subnet, connected to the same switch, hostnames are not involved, everything is done by static IP. When I log into vCenter I see the red triangles and a message that the HA agents something something, the message is cut off by the client interface (tbh, so disappointed that after all this time the client interface isn't fluid but fixed to a what I consider unworkable narrow width).

So, bottom line, I have no idea why the HA agents can't talk to vCenter in a timely fashion, and I am not aware of a good way to find out what the problem is. My HA cluster is basically hosed half the time. It doesn't affect the running VMs, but should there be a failure there would be an issue.

Any help would be much appreciated.

Edit: Actual it says time-out while communicating with the HA agent.
 
Last edited:
i would suggest modifying the cluster config, remove HA, then try re-enabling. the HA agent does this from time to time, in my experience.

additionally, I would go into the clusters events/tasks, or alarms, and look for the expanded error information.
 
Won't work. you need hostnames to resolve and IP addresses for HA. Give them hostnames, make an /etc/hosts file for windows and all of them with all hosts and VC, and then go from there. :)
 
lopoetve said:
Won't work. you need hostnames to resolve and IP addresses for HA. Give them hostnames, make an /etc/hosts file for windows and all of them with all hosts and VC, and then go from there. :)
Click to expand...

What he said. DNS resolution (and hosts files as a necessary backup) are an absolute must with ESX. I had similar issues until I did the same. I now maintain hosts files on all hosts and the vCenter server. This will ensure everything works even if they lose connectivity to the DNS server.
 
and the /etc/hosts files fix all the annoying niggles that HA loves to pick up about your DNS :p
 
This makes me a sad panda, I wasn't going to run DNS on my non-routed subnet. Thanks for the help though, I'll see about installing DNS.
 
To be clear - if you have fully accurate hosts files, a DNS server is not necessary. At least with 3.5 it isn't (I ran for months this way). Name resolution is what is compulsory.

I had the same situation as you, no DNS near the cluster as my setup was detached from the greater network.
 
SpaceHonkey said:
To be clear - if you have fully accurate hosts files, a DNS server is not necessary. At least with 3.5 it isn't (I ran for months this way). Name resolution is what is compulsory.

I had the same situation as you, no DNS near the cluster as my setup was detached from the greater network.
Click to expand...

Bingo - and it'll work in 4 as well, although you have a few more files to track (vdr needs one)
 
Don't hold out on me guysh! Just how many files are there to track? ;)
 
well, vc needs one, vcb will need one if you're using it, your DC will need one if you have AD, every esx host will need one, VDR will need one if you're using it, and vima will need one if you're using it. That's all I can think of off the top of my head :p
 
still have host files. DNS is only good when it works - the hosts files guarantee it works
 
I finally had they quiet moment to look at the HA error message and here is what it says;

Reconfigure HA host <hostname was here>.
Cannot complete the configuration of the HA agent on the host. See the task details for additional information. Unable to contact a primary HA agent.
Click to expand...

The machines are on the same subnet, DNS works. How do I track down why it's unable to contact the primary HA agent? The machines are all connected to the same switch. I didn't create the host files yet.
 
Build host files yet? I never trust DNS for HA. It's required, but always seems to be too picky. build a nice little host file and reconfigure - bet it works.

DRS doesn't care about HA and vice versa. The two actually aren't related :)
 
The VMs are on a SAN, yes.

I understand what you are saying lopoetve, host files are a word around though, what's the underlying problem? My DNS server is 3 hops from my ESX servers, they are all located within the same data center, there's no latency, there's no network issue that could cause frequent loss of connectivity to the DNS servers. It should just work, especially since I can't seem to find any reference to creating host files in the docs.
 
ESX/vCenter also seems to be case dependent sometimes. I try to make sure all host names are lowercase everywhere and that fixed problems I've had with name resolution. At least it did in 3.5.
 
Thuleman said:
The VMs are on a SAN, yes.

I understand what you are saying lopoetve, host files are a word around though, what's the underlying problem? My DNS server is 3 hops from my ESX servers, they are all located within the same data center, there's no latency, there's no network issue that could cause frequent loss of connectivity to the DNS servers. It should just work, especially since I can't seem to find any reference to creating host files in the docs.
Click to expand...

The underlying issue is that HA requires ~letter perfect~ forward and backward name resolution within a very very limited timeframe from your dns server. If it takes too long, if a letter is capped, if long/short/ip/reverse ip don't match up exactly like it expects, it throws a fit. And, on top of that, the agent is picky about things outside the dns spec - no capital letters, no all caps, hosts must be added by name not ip, etc. 4.0 is a bit more resilient.

You have to realize though - it's a legato agent, not VMware. It's purchased technology. :)

TBH, 90% of the installs I see have host files now. It just works. All you need is the ESX hosts and the VC machine and it'll be happy.

If you insist on using DNS... make sure nslookup long, short, and ip address all work. Make sure ARP works from the mac. Make sure /etc/sysconfig/network lists the right hostname, no capitals ANYWHERE, make sure DNS is fast, hosts are added via name not IP, make sure you can ping all of them, only one SC on a subnet, all of this from the command line.
 
ND40oz said:
ESX/vCenter also seems to be case dependent sometimes. I try to make sure all host names are lowercase everywhere and that fixed problems I've had with name resolution. At least it did in 3.5.
Click to expand...

Correct. DNS spec says it shouldn't matter. Legato agent says it does :(.
 
So that's something where the ESXi deployments are really hosed then? Since there's no good way to add a hosts file to ESXi? Meaning those who deploy ESXi instead of ESX but still register it to run HA and whatever else there is.
 
You don't need a host files, DNS isn't hard, just have it working like it's supposed to be and you're fine. I've never used a host file and I have 3 different vm clusters running with HA configured for all of them. And it's not like they're all using the same DNS servers since they're in different data centers spread throughout the U.S. Just need to make sure you have DNS working properly.
 
Thuleman said:
So that's something where the ESXi deployments are really hosed then? Since there's no good way to add a hosts file to ESXi? Meaning those who deploy ESXi instead of ESX but still register it to run HA and whatever else there is.
Click to expand...

Check the things listed in my post then :) One of those is missing something. Probably shortname or IP reverse lookup, in my experience.

DNS will do the job, if you have it perfect. It's just picky about everything.
 
There has to be some other issue then, my DNS was perfect, really.
I know I said earlier in this thread that I didn't have DNS set up, which was true for the internal network I had set up, but the DNS for the institutional network is perfect either which way.

However, my SAN is on the internal network and didn't have DNS at that time yet, perhaps that had something to do with the error ..., I really wish they would use more descriptive error messages, or have a KB that explains possible causes for specific errors. I am stuck on a different unrelated error which is completely non-descript. Ah well...
 
There are better error messages in the AAM logs. VI Client isn't a logging utility, it's just a gui to make changes :)

Double check the DNS from the host - I bet you find a hole somewhere.
 
I've been working on this same issue this week. I called support and they couldn't help me out, said it was a "deployment issue". haha
Anyway, here's what I had. VC is in the production network. I've got two new DMZ host servers (ESXi 4) I need to manage in my production VC. DMZ does not have DNS, so I had to count on my local hosts files. I kept running into aam errors, but after digging deeper into the vpxa logs, I could see "unable to get Fully Qualified Domain Name". After running a "hostname -s", I could see that the host server was not resolving. Ohh ya, maybe it would help if I put the long and short name of the host server I was on into the hosts file. after that, "hostnames -s" works! I also put in the NAT ip in the hosts file of the VC in production with port 902 open.
Within VC, I configure the "managed IP address" as the NAT address in the DMZ. I now connect my host servers to the cluster. Once they are in my new DMZ cluster, I configure HA options. Under advanced settings I set das.allowNetwork0 to the appropriate network. I now turn on HA. Everything now looks good, but the work is not over yet! On the host servers, I go back and configure <preserveServerIp>true</preserveServerIp>. I now go back to my VC and take out the managed IP address. If you leave this in, production host servers get disconnected when you try to manage the host. But now that I've taken this setting out, my DMZ hosts are fine because I preserved the NAT ip for VC in the vpxa.cfg file.
Hope this helps someone else. :p
 
We have plat support. The VMware support guys over at Dell just threw their hands up in the air when they didn't want to deal with it anymore.
 
Alright, so this does seem like a never-ending story.
I am now using hosts files, and it worked fine for a few days and then it hit the "HA agent has an error : Time-out while communicating with the HA agent" again. It also doesn't appear to recover from that error, the error condition can only be fixed by going to "Reconfigure for VMware HA".

Perhaps it's time to open an SR on this after all.

EDIT: The issue is particularly peculiar as the time-out message also appeared on the physical host which houses the vCenter VM. In theory that communication should never time out as it doesn't actually leave the physical server.
 
Last edited:
Hi,

I posted here because I have similar problem in my network.

I have vCenter at Intranet and two ESXs in DMZ wich must be in cluster. ESXs MGMT interface (interface through which vCenter communicate with ESX) is also in DMZ, but at another FW.

vCenter has IP 10.10.1.10/24, and ESXs have IPs 192.168.10.10/24 and 192.168.10.11/24. For test purposes all IP traffic is allowed trough FW in both directions.

For security reasons we are using NAT, so vCenter sees ESXs like 172.16.20.10 and 172.16.20.11, and ESXs see vCenter like 172.16.10.10.

There is network connectivity between vCenter and ESXs and between ESXs.

We can add ESX to vCenter. First we had problem with loosing connectivity in vCenter, but we found solution here: http://kb.vmware.com/selfservice/mi...nguage=en_US&cmd=displayKC&externalId=1010652

After that, ESXs are stable in vCenter. When we tried to clone or copy VM from one ESX to another, we can't do that. Process start and stop, after few minutes notification about connectivity problem shows. We also tried to make HA cluster, but same thing happend.

After customization host file of each ESX things got better. If we at ESX host file for address another ESX write real address (192.168.10.xx) clone/copy works, but HA don't. If we for address of another ESX write NAT adress (172.16.20.xx) HA works but clone/copy don't.

I attach example picture.

Have anybody have a similar problem?

[img=http://img109.imageshack.us/img109/8871/examplec.th.jpg]

Direc link for picture:

http://img109.imageshack.us/img109/8871/examplec.jpg
 
You must log in or register to reply here.
Back
Top