When I make a change to my scripts or do any type of configuration change that requires publishing to the agents (I have 29 of them), I will almost always get failures upon publishing. It's usually a mix. Some succeed. Some fail. However, there is a large difference between whether it is an agent manager initiated connection or an agent initiated. The former almost always succeeding and the latter being the one that frequently will fail. Of the 29 agents, 25 of them are of the agent-initiated connection that is frequently failing to publish.
I'll publish again and a couple more will succeed, most will still fail though. I'll publish again and maybe 1 more will succeed and the rest fail (the rest of the ones that were failing from the beginning that is... of course all the previously succeeded ones remain unchanged).
Sometimes I may be able to get all to successfully publish after a few tries but many times there will be a couple or even several (8 right now) that continue to fail no matter how many times I publish. But then maybe I'll try the next day and finally they'll publish.
Is this normal behavior? Does it just mean there's bad internet connection between the agents and agent manager? I would really like to fix this if it's in my power to do so, because having to publish repeatedly for days to get a crucial change to take hold is... problematic.
I'll open a ticket with support if needed but just checking first if anyone can confirm if this is just a network problem and if there is or isn't anything that can be done about it.
p.s. most of these agents are at vendor locations and I don't have direct access to the machines, but it is possible to use bomgar to connect with them
Answer by Steve M. ·
I did open a ticket regarding this issue and basically the conclusion was that it is a network problem.
So essentially there is no ultimate fix. But there was a suggestion that will potentially lessen the amount of failed publishes.
The suggestion was to use the IPs instead of the machine names or better yet use the fully qualified domain names. I'm talking about in the Active Agents tab. Double click on any of your agents listed there and in the form you can change the "Machine" field to the FQDN or just to the IP - the former being a little better.
For me I am still just using the machine names because despite the fact that it should be easy to get the needed information it unfortunately isn't. But if I ever do get this tried out, I will see how it goes for a few weeks and report back if it helps.
Troubleshooting tip: A handy tip in terms of a little troubleshooting - you can go to the agent webpage http://<machine name/ip>:9016 (username: "Vantage"; password: none) of an agent that the publish is failing for. Do this from the corresponding agent manager. If the page loads then the communication is fine (make sure you have compatibility mode set for the site). If the agent webpage doesn't load, there is some issue in terms of the network connection between the two.