I have also never seen people use a ZK solr.xml, and I see the solr.xml as a node-config file. I'd be very happy to only support file-system solr.xml loading.
On Wed, Sep 21, 2022 at 9:11 AM Jan Høydahl <jan....@cominvent.com> wrote: > Hi, > > Since 9.0 Solr can start with an empty SOLR_HOME as it will use defaults > in their place < > https://solr.apache.org/guide/solr/9_0/upgrade-notes/major-changes-in-solr-9.html>: > "Solr no longer requires a solr.xml in $SOLR_HOME. If one is not found, > Solr will instead use the default one from $SOLR_TIP/server/solr/solr.xml." > > So this motivation for storing solr.xml centrally is no longer valid. > > I agree with David's comment on the JIRA that this is also a question > about what we want solr.xml to be conceptually - a node-config file, or a > cluster-config file. We have other locations for cluster-wide > configuration. Today I think solr.xml is a mix of the two. A value like > "zkClientTimeout" could be cluster-wide while "host" and "hostPort" are of > course node-local. > > I'm also thinking about rolling upgrade scenario. So you do a rolling > upgrade from solr 8.11 to 9.0. For the 9.0 nodes you need different > configuration, maybe even radically different XML. Today most values in > solr.xml is sourced from Java System Properties, which is also a strong > hint that this is per-node. > > Jan > > > 20. sep. 2022 kl. 22:50 skrev Shawn Heisey <elyog...@elyograg.org.INVALID > >: > > > > On 9/20/22 14:18, Jan Høydahl wrote: > >> It has been possible to load solr.xml centrally from zookeeper for a > long time. > >> However, I'm considering deprecating and removing this feature. > >> Please see https://issues.apache.org/jira/browse/SOLR-15959 for > motivation. > >> > >> My question to the users list is thus - are you loading solr.xml from > zookeeper? > >> And if yes, why is that capability important for you - i.e. could you > not configure it per node? > > > > I have done very little with SolrCloud myself. I converted my tiny > little install to cloud with embedded zk, just the one server, one > collection and one core. I do not have solr.xml in ZK. I did this so I > have access to whatever functionality is cloud-only, should a need ever > arise. I fiddle with that install sometimes to try and answer support > questions. Rebuilding the index only takes about ten minutes, so if I > screw something up I just restore the working config, delete the data > directory, restart, and reindex. > > > > I can see a lot of value in being able to fire up a Solr node with only > /etc/default/solr.in.sh being provided. The solr home can then be > provided completely empty and the node will start, as long as solr.xml is > in ZK. > > > > One thing I think we should do is make it so that Solr starts with some > defaults, if solr.xml is not found at all. We can then bikeshed about what > the defaults should be. > > > > Making it possible for cloud mode to start without solr.xml might remove > most people's need to have it live in ZK. It would make things easier on > docker users ... they would be able to attach a completely empty volume for > the solr home and Solr would start. They might then go back and add a > solr.xml to provide custom settings. > > > > Thanks, > > Shawn > > > >
