| Summary: | Massive memory spikes from solr | ||
|---|---|---|---|
| Product: | [ECD] Orion | Reporter: | John Arthorne <john.arthorne> |
| Component: | Server | Assignee: | John Arthorne <john.arthorne> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | P3 | ||
| Version: | unspecified | ||
| Target Milestone: | 0.3 M2 | ||
| Hardware: | PC | ||
| OS: | Windows 7 | ||
| Whiteboard: | |||
|
Description
John Arthorne
The problem was the index purge job did a query for "everything" (in chunks of 1000 results at once), and then checked if each file was still present. By default solr returns all fields, including the contents of the files. So, we had the contents of 1000 files in memory during the query. In this case all we needed was the ID of each solr document, which vastly reduces the memory overhead. With this fix I am seeing a spike of 5MB rather than 300MB during index operations. http://git.eclipse.org/c/orion/org.eclipse.orion.server.git/commit/?id=7a062e8c082fa90b42da20353f00c203a7790b91 |