So, I reworked how caching works a bit, and now it seems quite a bit faster and stable for searches as well - although far from instant.
Ramakrishna - please let me know if you see the same thing! best regards, Anton On Sat, 18 Oct 2025 at 08:57, Anton Hägerstrand <[email protected]> wrote: > Hi Ramakrishna! > > I run Blunders.io. The root issue here is quite simple as far as I can > tell - the search profiles are just much larger. > > Blunders is not really built for the types of large JFR (Flight Recorder > profiling files) uploads that lucene-bench does, and it struggles a bit > with them. The main use of Blunders is a real time profiling platform, > where many much smaller profiles are used. Lucenebench instead uploads one > big .jfr per benchmark run (e.g. > https://blunders.io/jfr-demo/searching-2025.10.15.18.03.41/jvm_info). I > have not, so far, spent that much time optimizing for this use-case. > > I will have a look at if I can improve things somewhat - but a workaround > that works decently well is just knowing that Blunders has internal caches, > and that if you try reloading one of the profiles that timed out it might > very well have succeeded later. > > best regards, > Anton > > On Fri, 17 Oct 2025 at 15:12, Chilaka Ramakrishna <[email protected]> > wrote: > >> Hi all, >> >> I’m going through profiles on Blunders.io <https://blunders.io>. The >> indexing profiles load fine for me and I can click around and view the >> visualizations (very cool!). >> >> However, when I try to open search profiles, they consistently time out. >> For example: >> https://blunders.io/jfr-demo/searching-2025.10.15.18.03.41/jvm_info >> >> There was a similar issue earlier: >> https://lists.apache.org/thread/rxlj8zn0jlr3r4y0zg4xzn4nhwcg44wv (Thanks >> Mike!) >> >> Has anyone else been seeing the same problem recently? >> >> Best regards, >> Ramakrishna >> >
