[
https://issues.apache.org/jira/browse/SOLR-18318?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18099089#comment-18099089
]
Yash Goswmai commented on SOLR-18318:
-------------------------------------
Hi [~epugh] ,
I have submitted a PR for SOLR-18318.
The fix initializes `NodeSystemResponse.remoteNodeData` to prevent the NPE
during v2 `/api/node/system` proxying with the `nodes` parameter.
I have also added a SolrCloud test (`GetNodeSystemInfoSolrCloudTest`) covering
the failing scenario.
PR: [click
|https://github.com/Yashgoswami-ds/solr/pull/new/SOLR-18318-fix-node-system-npe]
> GetNodeSystemInfo (v2 `/api/node/system`) throws NPE when the nodes param is
> set
> --------------------------------------------------------------------------------
>
> Key: SOLR-18318
> URL: https://issues.apache.org/jira/browse/SOLR-18318
> Project: Solr
> Issue Type: Bug
> Components: Admin UI, v2 API
> Affects Versions: 10.0
> Reporter: Eric Pugh
> Priority: Major
> Labels: pull-request-available
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Calling {{GET /api/node/system?nodes=<nodeName>}} in SolrCloud mode
> (requesting system info for one or more remote nodes via the v2 API) throws a
> 500 with a NullPointerException, instead of returning aggregated per-node
> data the way the v1 equivalent ({{{}/solr/admin/info/system?nodes=...{}}})
> does.
> Note: this is a SolrCloud-mode issue where node resolution succeeds and
> proxying actually starts - it is a different bug from the
> null-{{{}ZkController{}}} NPE in standalone mode tracked by SOLR-18317,
> though both live in the same
> {{{}RemoteRequestProxy{}}}/{{{}V2SolrRequestBasedProxy{}}} machinery
> introduced by SOLR-16738.
> h3. Steps to reproduce
> A single-node SolrCloud cluster is enough - proxying to itself still
> exercises the code path:
> {code:java}
> bin/solr start -c
> NODE=$(curl -s
> "http://localhost:8983/solr/admin/collections?action=CLUSTERSTATUS&wt=json"; \
> | python3 -c "import json,sys;
> print(json.load(sys.stdin)['cluster']['live_nodes'][0])")
> # v1 -- works fine, returns {"<node>": {...fullSystemInfo...},
> "responseHeader": {...}}
> curl -s "http://localhost:8983/solr/admin/info/system?wt=json&nodes=$NODE";
> # v2 -- throws 500 NullPointerException
> curl -s "http://localhost:8983/api/node/system?nodes=$NODE";
> {code}
> h3. Actual result
> HTTP 500:
> {code:json}
> {
> "responseHeader": {"status": 500, "QTime": 77},
> "error": {
> "msg": "Error occurred while proxying to other node",
> "errorClass": "org.apache.solr.common.SolrException",
> "metadata": {"root-error-class": "java.lang.NullPointerException"}
> }
> }
> {code}
> Stack trace bottoms out at:
> {noformat}
> org.apache.solr.handler.admin.api.GetNodeSystemInfo$1.processTypedProxiedResponse(GetNodeSystemInfo.java:80)
> org.apache.solr.handler.admin.proxy.V2SolrRequestBasedProxy.processProxiedResponse(V2SolrRequestBasedProxy.java:77)
> org.apache.solr.handler.admin.proxy.RemoteRequestProxy.bulkProcessResponses(RemoteRequestProxy.java:96)
> org.apache.solr.handler.admin.proxy.RemoteRequestProxy.proxyRequest(RemoteRequestProxy.java:88)
> org.apache.solr.handler.admin.api.GetNodeSystemInfo.proxyToNodes(GetNodeSystemInfo.java:83)
> org.apache.solr.handler.admin.api.GetNodeSystemInfo.getNodeSystemInfo(GetNodeSystemInfo.java:57)
> {noformat}
> h3. Root cause
> {{solr/api/src/java/org/apache/solr/client/api/model/NodeSystemResponse.java:32}}
> declares:
> {code:java}
> public Map<String, Object> remoteNodeData;
> {code}
> with no initializer and no constructor that populates it.
> {{{}GetNodeSystemInfo.java{}}}'s {{proxyToNodes}} (lines 70-83) creates a
> fresh response and its anonymous {{V2SolrRequestBasedProxy}} override does:
> {code:java}
> // GetNodeSystemInfo.java:78-81
> public void processTypedProxiedResponse(String nodeName, NodeSystemResponse
> proxiedResponse) {
> response.remoteNodeData.put(nodeName, proxiedResponse); // NPE:
> remoteNodeData is null
> }
> {code}
> The same unguarded {{.put()}} also exists in
> {{NodeSystemResponse.setRemoteNodeResponse()}} (line 41, the Jackson
> {{{}@JsonAnySetter{}}}), which would NPE identically during deserialization
> of a response with unrecognized extra fields.
> h3. Suggested fix
> Initialize the field at declaration:
> {code:java}
> public Map<String, Object> remoteNodeData = new LinkedHashMap<>();
> {code}
> ({{{}LinkedHashMap{}}} for stable node ordering, consistent with other
> response maps in this codebase; plain {{HashMap}} also resolves the crash if
> ordering isn't a concern.)
> h3. Impact
> Blocks migrating {{{}solr/webapp/web/js/angular/controllers/cloud.js{}}}'s
> "Cloud -> Nodes" admin UI page from the v1 {{System.get(
> {nodes: ...}
> )}} call to the v2 {{SystemV2.getNodeSystemInfo(
> {nodes: ...}
> )}} equivalent.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
