kltm
commented
2 years ago Okay, I'm unsure where the "bad" SPARQL was, but we finally got this to barely work at 10m and 32GB (although monitoring the system indicates that 16GB may be sufficient). I've increased the time limit to 20m now just to make increase the likelihood of future success.
Made a copy of today's successful attempt at files.20220526/ in case there are future such problems, so we have a quick rollback (until we can get this into the main pipeline).
This done, the /go-cam/ site takes quite a while to load. It may be good to add a banner letting people know that it will eventually work. This may be an opportunity to revisit the site overall and think about what we want for it.
Tagging @thomaspd @dustine32
Recently, after the last release that added the imported data, the experimental go-cam site JSON builder (https://github.com/geneontology/pipeline/blob/issue-265-go-cam-products/Jenkinsfile) started failing. Specifically, the file gocam-gps.json only contains:
Some digging showed that the error occurred at exactly 30s. This is the embedded blazegraph's natural timeout. It seems reasonable that the increased gp count/query overhead is now timing out this one query.