[
https://issues.apache.org/jira/browse/SOLR-18549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18129907#comment-18129907
]
Eric Pugh commented on SOLR-18549:
----------------------------------
I probably need to consider standalone more.. In my world view, it's always
SolrCloud, so I don't think about apis that have a keyword like nodes even
making sense in standalone.
I am surprised that these end poitns blow up, as I thought they are exposed in
the Solr Admin UI?
I suppose, if they aren't supported in standalone, they shouldn't blow up, they
should offer a graceful error?
> Standalone node throws NPE for nodes or node parameters in proxy-capable
> admin endpoints
> ----------------------------------------------------------------------------------------
>
> Key: SOLR-18549
> URL: https://issues.apache.org/jira/browse/SOLR-18549
> Project: Solr
> Issue Type: Bug
> Reporter: Nick Shanin
> Priority: Major
>
> 🤖 _AI text below_ 🤖 _(posted on behalf of Nick Shanin)_
> Related: SOLR-18317 (the Admin UI half of the same report), SOLR-15823 (open;
> V2 logging broadcast).
> The {{nodes}} and {{node}} parameters on a standalone node (not running in
> SolrCloud) fail with a
> NullPointerException, or with a 500 error that wraps one, in four endpoints:
> * {{/admin/info/logging}} (v1), for reads and for {{set}}
> * {{/admin/info/system}} (v1)
> * {{/api/node/system}} (v2)
> * {{/admin/metrics}} (v1), with {{node}}
> The {{nodes}} and {{node}} parameters on a standalone node (not running in
> SolrCloud) fail with a NullPointerException, or with a 500 error that
> wraps one, in four endpoints:
> * {{/admin/info/logging}} (v1), for reads and for {{set}}
> * {{/admin/info/system}} (v1)
> * {{/api/node/system}} (v2)
> * {{/admin/metrics}} (v1), with {{node}}
> For example, {{GET /solr/admin/info/logging?nodes=all}} fails. A {{set}}
> request with {{nodes=all}} applies the level change on the node and then
> fails the same way. The shared proxy code
> (RemoteRequestProxy.validateNodeNames and GenericV1RequestProxy.shouldProxy)
> reads the ZooKeeper controller without checking that one exists. The Admin UI
> no longer sends {{nodes=all}} in standalone mode, but other clients, older
> UIs and direct calls still get the error.
> Expected behavior:
> * {{{}/admin/info/logging{}}}, {{/admin/info/system}} and
> {{{}/api/node/system{}}}: {{nodes=all}} is answered by the node itself. Other
> {{nodes}} values return 400 with a message that the parameter requires
> SolrCloud mode.
> * {{{}/admin/metrics{}}}: {{node=all}} and explicit node names return 400
> with the same kind of message.
> Outside this ticket: the V2 logging endpoint
> ({{{}/api/node/logging/levels{}}}). SOLR-15823 gives it a 400 for {{nodes}}
> in standalone mode. Open question: should the V2 logging endpoint answer
> {{nodes=all}} locally too, for consistency with v1? This ticket does not
> decide it.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
