Well, if you're rebooting the machines, there's not much to do except do them one at a time ;).
If you have a maintenance window where the service can be unavailable for a time I'd 1> stop indexing 2> insure any in-flight updates have been committed (just issue a commit to each collection) 3> just shut them all down gracefully. 4> bring the nodes all back up and start ZooKeeper on all three (assuming you're not using embedded ZK) 5> bring up my Solr nodes one by one. If you can't have the service unavailable, you don't have much choice but to do the nodes one at a time. And if you're not running embedded ZooKeeper I'd also start ZK first before starting Solr. I'd still stop indexing if I could, but that's not essential. You'll have a little less work if you down the Solr node with the most leaders _last_, but that's not essential either, it'd just avoid a bit of leader election overhead. Best, Erick On Sun, Dec 16, 2018 at 4:58 AM Moshe Recanati | KMS <mos...@kmslh.com> wrote: > > Hi, > > We've SolrCloud servers configuration with the following topology. > > 3 servers in which each server is zookeeper and Solr search. > > We would like to do maintained restart of the servers. > > Is there any specific guidelines on how to do the restart? > > > > Thank you > > > > Regards, > > Moshe Recanati > > CTO > > Mobile + 972-52-6194481 > > Skype : recanati > > More at: www.kmslh.com | LinkedIn | FB > >