[SOLR-14443] Make SolrLogPostTool resilient to unexpected requests

• Type: Improvement
• Status: Closed
• Priority: Trivial
• Resolution: Fixed
• Affects Version/s: 9.0
• Fix Version/s: 8.6, 9.0
• Component/s: scripts and tools
• Assignee: Jason Gerlowski
• Reporter: Jason Gerlowski
• Created: 28/Apr/20 18:57
• Updated: 15/Jul/20 15:09
• Resolved: 22/May/20 14:56

Description

When SolrLogPostTool parses log messages corresponding to incoming requests, it sets various predefined fields based on the parameters on the request. e.g. it sets a rows_i field, a wt_s field, and so on.

This logic works for most requests, but if the log-parser encounters requests with multiple of these params (e.g. rows), it will blithely add them to the SolrInputDocument, and error out when Solr rejects the eventual update request because it is attempting to put multiple values into a single-valued field.

We can do two things to fix this.

1. Make SolrLogPostTool's "posting" code resilient to individual update failures. It doesn't make any sense to crash the entire posting routine just because one batch (or one log message) was malformed.
2. Tweak the field parsing logic to be more resilient to the specific "redundant query params" case I encountered specifically here.

Issue Links

• links to: GitHub Pull Request #1525