graphite-dev team mailing list archive

[Cliff Pracht <question227319@xxxxxxxxxxxxxxxxxxxxx>]

Question #227319 on Graphite changed:
https://answers.launchpad.net/graphite/+question/227319

Cliff Pracht posted a new comment:
Thanks for the info Jason.

Our setup is a bit different.  The whisper files are on a (clustered)
NFS file share (IE Netapp FAS Clustered pair).  We have 3 Carbon Cache
Nodes which read/write to share.   We have an Apache node that is only
used for rendering graphs and dashboards. A single Carbon Relay process
currently relays metrics using the Rules-based mode.

What I was thinking to address some of the shortcomings of this setup is
to do the following:

The whisper files would remain on clustered NFS file share and 3 carbon
cache nodes would write to the share.  I would like to add an additional
apache node for rendering and put that behind an F5 VIP;  spin up an
additional Carbon-relay process,  change write algorithm to Consistent-
Hash mode and put those behind a rule. Not sure if I can have 2 active
Carbon-relays with round robin setup OR would need to setup an F5 rule
for persistence, so that it doesn't thrash.

So here are the paths depending on if you are reading metric data or
writing:

Reading metrics-> VIP -> apache1|apache2 -> Whisper over NFS(ro)

Writing metrics -> VIP -> CarbonRelay1|CarbonRelay2 ->
CarbonCache1|CarbonCache2|CarbonCache3-> Whisper over NFS(rw)

-- 
You received this question notification because you are a member of
graphite-dev, which is an answer contact for Graphite.