Monitoring Splunk

loadjob performance

Path Finder

What determines the performance of loading the artifacts of a savedsearch? I have a job which ran a savedsearch, and it has 70,000 results. Doing a 'loadjob' on the sid of that job takes 10 seconds. What could cause it to take so long?

Tags (2)
1 Solution

Champion

loadjob streams the contents of the dispatch directory for the given search into memory before executing. Thus, performance can vary based upon the size of the dispatch directory and the IO performance of your disks.

If you are not using 5.0.2 or later, then you may find performance improves if you upgrade. Before 5.0.2, loadjob used to load the dispatch directory for the search into memory all at once. Thus, using loadjob for a search with a 1 GB dispatch directory would use up 1 GB of memory. This could cause a significant reduction in performance. In 5.0.2, the contents are streamed so that the entire directory is not in memory all at once; this is much more performant.

View solution in original post

Champion

loadjob streams the contents of the dispatch directory for the given search into memory before executing. Thus, performance can vary based upon the size of the dispatch directory and the IO performance of your disks.

If you are not using 5.0.2 or later, then you may find performance improves if you upgrade. Before 5.0.2, loadjob used to load the dispatch directory for the search into memory all at once. Thus, using loadjob for a search with a 1 GB dispatch directory would use up 1 GB of memory. This could cause a significant reduction in performance. In 5.0.2, the contents are streamed so that the entire directory is not in memory all at once; this is much more performant.

View solution in original post