Well. I’m kinda aware about the stats :) But I believe they are not a good way for LB. The issue is LB requires a simple response OK and cannot do extended parsing.
I see following cases what can happen in production: 1) Server is alive - riak is stopped 2) Server is down completely 3) Server is alive, riak is alive but it’s not a part of a cluster. 4) Server is alive, riak console is alive but backend(bitcasc for example) is down. 5) What’s else? Honestly I’m considering to build a special Web service that will run together with Riak and give it status. So far I was not be able to find a standard way using Riak capabilities to cover all 4 possible scenarios. Thank you, Konstantin. On Feb 2, 2014, at 9:33 AM, Roberto Aloi <[email protected]> wrote: > Hi Konstantin, > > Riak exposes a number of metrics - such as counters and histograms - that you > can use to monitor your nodes' health. > Metrics are available through the HTTP API via the /stats endpoint. > For details, see: > http://docs.basho.com/riak/latest/ops/running/stats-and-monitoring/ > > Cheers, > > Roberto Aloi > @robertoaloi > > > > On Sun, Feb 2, 2014 at 6:23 PM, Konstantin Kalin <[email protected]> > wrote: > Sorry jumping into the discussion. What is a best way to monitor Riak node > health? Most loadbalancer uses HTTP request to check if a node is alive. > > Currently we use Riak Java client to load balance requests to Riak. The issue > is if a node gets removed or added all Java servers need to update > configuration to reflect the changes. It’s kinda annoying for operation team. > So we started thinking about putting LB between Riak cluster and Java > clients. > > Thank you, > Konstantin. > > > On Feb 2, 2014, at 8:25 AM, Gideon de Kok <[email protected]> wrote: > >> Mark, >> >> Your throughput of your total cluster will still be dependent on the >> throughput on that specific node: >> It can coordinate requests, but stays responsible for the actual response to >> the client. >> >> Load balancers not only help to increase performance by fanning out the >> requests to multiple servers, >> it will also increase the availability of your cluster in terms of client to >> server connections. >> >> Cheers, >> On 2 Feb 2014 at 17:20:28, markrthomas ([email protected]) wrote: >> >>> Hello >>> >>> Any node in a Riak cluster can coordinate requests for any other Riak node >>> that is responsible for the data you're reading/writing. >>> >>> What is the advantage of fronting a Riak cluster with a load-balancer? >>> >>> Thanks >>> >>> Mark >>> >>> >>> >>> >>> -- >>> View this message in context: >>> http://riak-users.197444.n3.nabble.com/Load-balancer-tp4030414.html >>> Sent from the Riak Users mailing list archive at Nabble.com. >>> >>> _______________________________________________ >>> riak-users mailing list >>> [email protected] >>> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com >> _______________________________________________ >> riak-users mailing list >> [email protected] >> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com > > > _______________________________________________ > riak-users mailing list > [email protected] > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com > > > > > -- > Roberto Aloi > --- > Website: http://roberto-aloi.com > Twitter: @robertoaloi
_______________________________________________ riak-users mailing list [email protected] http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
