Thanks a lot, Peter. I have got a fair idea now..
Just started to play with the standalone setup, in a single lab server.
Will come up with more questions, once I add more servers,etc..
On Mon, Dec 17, 2012 at 3:54 PM, Peter Haag <phaag@...:
> Hi Senthil,
> On 11/26/12 21:34, Senthil_Veni wrote:
> > Hello
> > First of all thanks to the opensource world & all the folks contributing
> to its success.
> > This is my first, hopefully a redundant question.
> > We are building a NetFlow Proof of Concept setup for a client and
> wondering if we can separate out the NF daemons into
> > multiple servers, for best performance & scalability.
> > (In production there are approx 500+ L3-Switches & routers with
> multicast,sending netflow v5 & v9 @ 12 to 15 million
> > flows per minute.)
> Note: NfSen as it is for now was not built with separation in mind. The
> solution below is all theory - I don't know
> if this really works, as I never did it. However, it could work :)
> > We are thinking abt the following setup for Proof of Concept:
> > -- 2 servers each with their own NFCapd daemons & NFCapd Files in a
> common NFS mount point in NAS.
> > -- 1 server with NFDump, NFProfile with access to NFCapd Files
> mount point.
> > -- 1 server with NFSen with access to NfCapd files & Live RRD files.
> > -- Tester / simulators for flows up to 10 million FPM.
> > Is this do-able? Any reference docs or hints on Distributed NFDump/NFSen
> would help?
> 1. Install nfdump on each of the collectors servers. You would need to
> setup NFS and the directories accordingly on
> those servers.
> 2. Install nfdump/NfSen on the "NfSen" server". Make sure NFS mountpoints
> match everywhere.
> Note: You can not have multiple of these, as they compete for profiling
> and expiring.
> 3. On The NfSen Server configure all sources as you would on a single box.
> Exception - set collectors port to 0.
> This prevent to start up a collector on the NfSen box.
> 4. You need now to start on each collectors bos all nfcapd processes
> manually exactly as they would run on a single
> /usr/local/bin/nfcapd -w -D -p <port> -u <user> -g <www-group> -B
> 4000000 -S 2 -P <pid file> -z -I <ident> -l
> 5. Make sure user/group IDs match on each box.
> 6. Make sure the clocks run synchronous on each box, also check the
> On The NfSen box, you now can setup profiles look at the graphs and flow
> files. This is most likely the heavier loaded box.
> Make sure you have an appropriate number of profilers configured.
> 7. Keep your fingers cross
> - Peter
> > --Senthil
> > (Note : This is just a PoC. In production, we will add more servers for
> NFCapd & look at better I/O performing storage
> > solution)
> > Monitor your physical, virtual and cloud infrastructure from a single
> > web console. Get in-depth insight into apps, servers, databases, vmware,
> > SAP, cloud infrastructure, etc. Download 30-day Free Trial.
> > Pricing starts from $795 for 25 servers or applications!
> > http://p.sf.net/sfu/zoho_dev2dev_nov
> > _______________________________________________
> > Nfdump-discuss mailing list
> > Nfdump-discuss@...
> > https://lists.sourceforge.net/lists/listinfo/nfdump-discuss
> Be nice to your netflow data