Closed dnil closed 11 months ago
Also the general case report is fairly slow again. Hmm..
Given how unpredictable this seems, I think we may mostly be dealing with fallout from a large index size compared to system RAM again. We have stepped over the all-indexes-in-memory bound since a little while, and will be at the mercy of certain optimisation whims.
That is not counting the extra dbs co-hosted. Most are smallish, but e.g. loqusdb-rd is sizeable.
We also have that mongod-backup-service for the "smallish" dbs, but that does include loqusdb-rd. Whenever that runs we will have a reshuffle of a sizeable part of active memory.
Also the general case report is fairly slow again. Hmm..
At least the report speed can be improved relatively easily perhaps?
Potential local fix: https://github.com/Clinical-Genomics/IT-issues/issues/733
Potential local fix: Clinical-Genomics/IT-issues#733
That's cheating!
Ok, we are not in the clear yet - both the case and variantS views make slow queries, but a qualitative investigation says this and the past few optimisation PRs helped a bit. I'm closing this for now based on completion of https://github.com/Clinical-Genomics/IT-issues/issues/733, and we can keep working on query optimisations e.g. https://github.com/Clinical-Genomics/scout/issues/4153 and possibly more batch processing for e.g. collapsing identical variant documents.
Certain views have gone a bit slow lately - case, variantS come to mind. I haven't dug deep on this, but there are some slow queries I don't quite recognise. They don't look particularly nasty, like pollscan on cases: even if has grown quite a bit, it is still not huge, but let's see what we can do.