How agent selects a SRS when he get multiple primay server?
Sorry for my english
I ask you to figure out the selection by an agent of several servers SRS when he get multiple primay server
From this place "F-Secure Offload Analyzer Agent> Settings> Connection" ("Primary Servers" and "Backup Servers"), we can specify mulitple IP
On the other hand, how is an agent selected on the IP of the main server?
- Does it scan the list sequentially?
- Does it use the "connection time" field to determine if the server is available?
1) If multiple Primary Servers, they are selected by response time.
2) Backup Server same, but only after ALL primaries failed.
3) It is highly recommended to have two primaries equally configured.
There is not much network load here and as long as the servers are in the same backbone you should not worry about anything.
If, in very small environments, you have two sites with servers on both ends and you want the one to be failover for the other you configure them as Primary and Backup and vice-versa.
In a standard datacenter environment where VMs move from one hypervisor to an other on the fly, you put two SRSs as primary. So that each can update while the other answers the requests.
My recommendation: add two primaries and observe, additional confiruration is rarely needed.5 Like
Thanks for your reply.
Let me "rephrase" in order to be sure that I have correctly understood:
-If we put 3 SRSs as primary, each agent will contact the three at the same time and based on response time from them, one (the fastest) will be selected: is it right?
Last point: when an agent has selected one SRS, he will keep the connexion to this one during a certain temps, but for how long? Another way, how long a agent decides to re-select a new one ?
While I can not answer those questions (@vad) from knowledge I can only report that I have never heard of a problem rising from the voting algorithm. Yes, one is selected but when and why that gets re-evaluated I do not know.
I assume that the SRS's answer must arrive within 5 sec (connection timeout) or the agent will re-check the timing. I also expect that to happen on a regular basis anyway.
Some other thought: in case we have 50 VDIs it would be wanted, that all connect to the same SRS. This way the cached results form the first VDI wil speed up answers to the other 49. OTOH if three SRS exist the first VDIs would populate the chaches and the remaining VDI might get a slightly better performance due to less response time.
But if you run a server farm with many differen setups,APPs, versions, OSs and Patches installed the caching might be less efficent anyway and better performance trou three SRSs would be your choice.
All in all the concept has never shown to be a bottleneck and usually speed up the VMs, not slow them down. And adding a fourth SRS is a matter of seconds and can be done with the snip of your fingers.
It would be interesting to learn what you fear and what is your setup?