graphite-dev team mailing list archive

[Bob A <question232334@xxxxxxxxxxxxxxxxxxxxx>]

New question #232334 on Graphite:
https://answers.launchpad.net/graphite/+question/232334

I'm working to find a HA configuration for graphite(graphite-web, carbon-cache) so that we can start to rely on the data being available in production with a given host failure.

I have carbon-relay setup to send all data(RELAY-METHOD = rules) to two servers running carbon-cache, and graphite-web.  These two hosts are behind a loadbalancer and graphite-web(port 80) requests are being loadblanced between the two.  On each host, I have CLUSTER_SERVERS in local_settings.py set to point to the other graphite-web instance.  

The trouble seems to come when one of the carbon-cache processes is not running OR does not have the entire data set for the metric I am trying to graph over the requested period.  I am seeing incomplete data when requesting metrics directly from the graphite-web host :80 that has the incomplete local data(or this will happen every other time when making the request from the VIP due to round robin lb'ing).  In this test case, the complete data set for the requested period of time is on the other graphite-web host listed in CLUSTER_SERVERS.  I am hoping there is a way to configure graphite to also look at the server listed in CLUSTER_SERVERS to mix the data even though it found the data(albeit incomplete for the requested time period) in the whisper db on the local filesystem. 

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