Can someone help me with a quick question before I go ahead and install some agent handlers?
I work for local authority which has 3 main sites (1000+ users) and around 20 smaller sites (1-100 users). For the smaller sites which have a local server I have installed the McAfee agent and upgraded them to SuperAgents so they are also repositories.
I am running ePO4.5 and have seen that Agent Handlers could offer better network bandwidth along with other things but can I install an agent handler on a server that is already a SuperAgent? If so during the install what port should I specify? Is it the console to server port?
Agent Handler should only be used as a scalability method to take communication load away from the ePO server as they require a high availablity connection to the ePO database server.
There is a common misunderstanding in how Agent Handler work as most of our customers assume that they can use an agent handler as a remote epo server with a slow connection to a site which is clearly not the case.
Additionally the full functionality of Agent Handler can only be used at the moment via McAfee Agent 4.5 (which by the way is not supported on Win2000 and below).
Just a side note, we DO support MA 4.0 with Agent Handlers. Its just that the 4.0 agent can only have one servers AH listed in the sitelist at a time. So it makes it difficult to maintain.
Also adding to the point of the AH will not save bandwith, this is because the AH will have to talk to the eposerver for package updates, and directly to the SQL server for policies, and tasks.
Now on the flip side, it would stream line the bandwidth to just one/two machines, and if you use the AH in the remote sites as a repository, it would be better on bandwidth than the machines updating from the Main epo server. So you may still want to use them, just depends on what you are trying to accomplish.
That sounds like what I want to do. What happens at the moment is that all the machines on the remote sites talk directly to the epo server (for policies and tasks) and their local file server (which is a SuperAgent) for updates. But am I right in saying that with an Agent handler installed on the local file server that the remote clients would talk to the agent handler and then that would talk back to the epo server (saving bandwidth). I agree that when the agent handler talks back to the epo sever that the bandwidth usage will be higher than normal as it has more data to send back (about all the remote clients) but it will be for a short amount of time where as with my current setup the is little bandwidth usage from the clients themselves but alot more of it as there are many clients.
One more thing. Is it OK to install an agent handler on a server that is already a SuperAgent or would there be no need for the server to be a SuperAgent anymore as the agent handler would do the same sort of role?
Deanon 11/20/09 4:09 AM
I'm afraid this isn't going to save you bandwidth - the AH needs a permanent connection to the SQL server. If reducing bandwidth is your goal, then agent handlers are not the solution, I'm afraid.
Perhaps we can approach this from the opposite direction: what exactly are we trying to achieve? Are we trying to reduce bandwidth between the remote sites and the main site?
I didnt realise that a permenant connection was needed. I thought it was going to be just as and when :-(
I thought I would try out the AH for 2 reasons really. One because it is a new feature and was advised to try them out by our resller and also because I have an intermittent problem with some SuperAgent sites where after the SuperAgent server reloads sometime the agent does not bind to its port and therefore I am unable to do wake ups and replicate to that site. Beacuase the agent is no longer listening clients ignore it as a repository and then have to come across the ADSL broadband line to another repository causing numerous problems with traffic and network speeds.
Maybe a permenant connection is not such a bad thing but I would need to know how much constant bandwidth we are talking about?
I'm not sure what the bandwidth would be, I'm afraid - it would depend on how many clients were on the site, how often they were contacting the AH, and so on. We haven't got a sizing guide specifically for ePO 4.5 yet, but you can use the existing ePO 4 sizing guide as a guideline. You can basically treat an agent handler as though it were an ePO 4 server.
Thanks for the info Joe. Its been real helpful. I think im going to give it a go on one or two sites to see what sort of performance I get.
One last thing, should I remove the SuperAgent functionality of the agent from the remote server first before making it a AH? Also when installing the AH and it asks for a port what port number do I use? Is it the console to server port I have already configured?
yes ... I would remove the SuperAgent functionality prior to install the AH on the node ... and ...
yes ... it is the console to server port which you will need to choose in the first installer mask ...