Use a load balancer. We’re in AWS, so we use an AWS ALB. If you don’t have a failure-tolerant load balancer implementation, the site has bigger problems than search.
wunder Walter Underwood wun...@wunderwood.org http://observer.wunderwood.org/ (my blog) > On Jan 11, 2021, at 10:15 AM, Dmitri Maziuk <dmitri.maz...@gmail.com> wrote: > > On 1/11/2021 11:25 AM, Walter Underwood wrote: >> There are all sorts of problems with the primary/secondary approach. How do >> you know >> the secondary is working? How do you deal with cold caches on the secondary >> when it >> suddenly gets lots of load? >> Instead, size the cluster with the number of hosts you need, then add one. >> Send traffic >> to all of them. If any of them goes down, you have the capacity to handle >> the traffic. >> This is called “N+1 provisioning”. > > Where do you send your solr queries? If you have an http server at an ip > address that answers them, that's a single point of failure unless you put it > on a heartbet'ed cluster ip. (I tend to prefer ucarp to pacemaker for that as > the latter is bloated and too cumbersome for simple active/passive setups, > but that's OT.) > > Dima
