Load balancing doesn't work...
... because the host returned by the collector is not correct.
In our case, we have two physical servers (preprodcoll01 and preprodcoll02 )with 5 collector instances on each one of them. The five instances on the first server are overloaded while there are no agents on the second server.
When we took a closer look, we noticed it was because the host name being returned by the collector seems to be a short version while our agents need the fully qualified name to connect.
For example, it returns "preprodcoll01" while it should return "preprodcoll01.mycompany.com" instead.
Is there a way we can force the collector to use a different host name(like the "external host name" configuration for the server)?
Thanks in advance!
Chris
Going to move this to the Dynatrace Open Q&A forum as this is currently posted in the general Community Open Q&A
Actually I think a network service as the collector should ALWAYS return the FQDN (if it can be detected safely) as you can never assume for proper short DNS resolution.
What you probably can do Chis is to make sure the "hostname" command on the machines returns the full qualified name instead of the short name on the machines...
Thanks for the info Reinhard! Is there no other way we can change this?
I mean, there is an option for the server but not for the collectors?!
I believe you can go to Settings/dynatrace Servers/Collectors/Advanced Configuration and add a bind address.
Doesn't the bind address bind the collector to only specific hosts?
That's not what Chris is looking for. If the first collector in a loadbalanced configuration tells the client to connect to "thishostname" but the agent can't resolve "thishostname" there should be an option to set that the colelctor should return the full qualified name "thishostname.domain.my".
I thought that was what the bind address did.
We had similar issues with EasyTravel and used the bind address to fix it:
Re: Can't get Collector Group to work on 6.0 with EasyTravel
But maybe I am missing something.
Thanks,
dave
Actually that might be a good workaround putting the FQDN in there and hoping it resolves locally to the collector as well. Otherwise I'd just put the IP adress in there and hope that the balanced collector forwards this to the agents.
hi everybody,
David is right, that's actually exactly what those bind addresses are for.
as taken from the doc (Collectors):
"The Collector name that is sent to the Agent for storage in its CollectorList file is the detected host name of the Collector. If an IP address (multiple NICs in the machine) or fully qualified domain name is required for the Agent to connect to the Collector, you can choose to configure a Collector Bind Address. This bind address will be sent to the Agent and populated in its CollectorList file. To configure this, switch from the Basic to the Advanced Configuration in the dialog."
HTH,
Christian
I am happy to report that bind addresses and changing the hostname both seem to have done the trick as my agents are slowly starting to connect to all my collectors and not only to only one.
Thanks everyone for the help!
Glad the bind addresses worked for you. I also saw something innovative recently where someone used a bind address of 127.0.0.1 for a Monitor Collector to (I assume) prevent regular agents from accidentally being pointed at it. I thought this was a nice practice. Has anyone else seen or done this? Does it seem like a good idea?
I'd say it's a very good idea indeed! This way, we are sure that no agents are going to connect to it (unless they are local).
