Re: heartbeat question

Artyom Khmelnitsky <[email protected]> Mon, 29 Mar 2010 00:13:31 -0400
Newsgroups gmane.linux.highavailability.ultramonkey
Organization AYKsolutions, LLC.
Message-ID <[email protected]>
HI Simon,

Thank you for your reply. This particular instance is a site that 
converts videos to mp3s and then allows the visitor to download them. It 
has some custom scripting on it and uses ffmpeg in some instances. The 
current single machine is being overloaded by too many ffmpeg instances, 
so the load needs to be split up. Ultramonkey will then have to 
check/monitor each end-server's load or connections and forward the user 
to the least busy node.

Does LVS and feedbackd come pre-packaged with the tar file for 
ultramonkey? I believe LVS does?


Simon Horman wrote:
> On Sun, Mar 28, 2010 at 06:34:09PM -0400, Artyom Khmelnitsky wrote:
>   
>> Hello,
>>
>> We are thinking of utilizing heartbeat for one of our customers. I have 
>> been reading up on the ultramonkey website, but was not able to find an 
>> answer to my question. How does heartbeat choose the backend node? Is it 
>> done randomly like with rrDNS or does it monitor the webserver's load 
>> and choose the one with the lowest?
>>     
>
> The real-server selection is done by LVS, which is configured
> by ldirectord through the ldirectord.cf file. They key is
> the choice of scheduler. Typically ultramonkey uses the wlc scheduler,
> which basically chooses the scheduler that has the lowest number of
> connections.
>
> A separate project, feedbackd, looked at the idea of monitoring
> the load on real-servers and by tweaking their weights in
> the LVS table it was able to send more connections to less loaded
> servers.
>
> Is there a particular scenario you have in mind which calls
> for scheduling based on server-load?
>
>
>   
_______________________________________________
Ultramonkey-users mailing list
[email protected]
http://lists.vergenet.net/listinfo/ultramonkey-users