Open zhaih opened 2 years ago
Big +1 to this.
Right now, luceneutil does not work "out of the box" without any local code changes due to the File name too long
error that jfr throws.
To reproduce:
python3 src/python/localrun.py -source wikimediumall -r
Create indices:
/local/home/jslowins/upstream_bench/lucene_bench_home/indices/wikimediumall.lucene_candidate.candidate.facets.taxonomy:Date.taxonomy:Month.taxonomy:DayOfYear.sortedset:Date.sortedset:Month.sortedset:DayOfYear.taxonomy:RandomLabel.sortedset
:RandomLabel.Lucene90.Lucene99.dvfields.nd33.3326M: now create
cd /local/home/jslowins/upstream_bench/lucene_bench_home/lucene_candidate/lucene/benchmark
log /local/home/jslowins/upstream_bench/lucene_bench_home/logs/baseline_vs_patch.wikimediumall.lucene_candidate.candidate.facets.taxonomy:Date.taxonomy:Month.taxonomy:DayOfYear.sortedset:Date.sortedset:Month.sortedset:DayOfYear.taxonomy:
RandomLabel.sortedset:RandomLabel.Lucene90.Lucene99.dvfields.nd33.3326M.log
WARNING: Using incubator modules: jdk.incubator.vector
[0.302s][error][jfr,startup] Could not start recording, not able to write to file /local/home/jslowins/upstream_bench/lucene_bench_home/logs/bench-index-baseline_vs_patch-wikimediumall.lucene_candidate.candidate.facets.taxonomy:Date.taxonomy
:Month.taxonomy:DayOfYear.sortedset:Date.sortedset:Month.sortedset:DayOfYear.taxonomy:RandomLabel.sortedset:RandomLabel.Lucene90.Lucene99.dvfields.nd33.3326M.jfr. /local/home/jslowins/upstream_bench/lucene_bench_home/logs/bench-index-baselin
e_vs_patch-wikimediumall.lucene_candidate.candidate.facets.taxonomy:Date.taxonomy:Month.taxonomy:DayOfYear.sortedset:Date.sortedset:Month.sortedset:DayOfYear.taxonomy:RandomLabel.sortedset:RandomLabel.Lucene90.Lucene99.dvfields.nd33.3326M.jf
r: File name too long
Error occurred during initialization of VM
Failure when starting JFR on_create_vm_3
See this comment, the auto generated index name could get overlong and JFR is complaining about it. Altho it can be easily overcame by just give the index a dedicated name, but anyway I think make the auto generated index name shorter is still necessary: right now we're trying to list the index spec as part of the name but that will make it extremely hard to open any file using this index name. Perhaps we can just use timestamp as generated index name, and then put the specs in some separated file inside index directory?