Hi, In Solr 9 the focus is on a rich Java API that will allow you to create your own placement plugin that can do pretty much anything. The affinity plugin tries to give a good OOTB experience for most use cases, but does not aim to be the one everyone uses. You may want to fork the affinity plugin and extend it with your needs. If you believe your extension is useful to a larger audience you may consider opening a PR with your changes.
Jan > 12. okt. 2022 kl. 16:05 skrev Jonathan Tan <jty....@gmail.com>: > > Hi! > > In Solr 8.6, I used the autoscaling configuration to set it up so that we > had 2 `nodeTypes`, one for indexing, the other for querying. > And our collections were set up to be TLOG + PULL replicas only, and we'd > use the autoscaling to place the TLOG on the indexing nodeTypes, and the > PULL on the querying nodeTypes. (Essentially replicating the older > master/slave setup of pre-SolrCloud) > > Now we've come to Solr 9, and I can't really find a clear way to do the > same thing. > > https://solr.apache.org/guide/solr/latest/configuration-guide/replica-placement-plugins.html#affinityplacementfactory > > I've looked through this, and it seems to imply that the > `collectionNodeType` configuration will allow me to place specific > collections on specific nodeTypes. But I can't find a way to place specific > replica types on specific nodeTypes... > > Is this mechanism no longer existing? > > Thank you > Jonathan
