https://www.javaadvent.com/2019/12/measuring-time-from-java-to-kernel-and-back.html It seems like nanoTime and timeAllowed may hit a performance bummer in a certain condition. Beware!
On Sat, Dec 17, 2022 at 12:21 AM Mikhail Khludnev <[email protected]> wrote: > Thanks Keisuke. > Adding users, and Walter. > I always thought that timeAllowed checks have a low footprint. > > I googled 6 y.o thread mentioning some low level specifics. > Walter, do you remember what was the outcome of those experiments? Can you > share your observations? > > https://solr-user.lucene.apache.narkive.com/FYXlrhMP/solr-performance-on-ec2-linux > > Beside that I suppose timeAllowed should be smaller than client side > request timeout to let server to return partial results before cutting > connection or so. Just my gut feeling. > > On Fri, Dec 16, 2022 at 4:17 PM Keisuke Isono <[email protected]> wrote: > >> Hi, Alessandro, Mikhail, Sekiguchi-san, >> >> I provide information about the timeAllowed parameter. >> We have not been able to look into the details, so this is only what we >> know. >> >> system information >> OS: AmazonLinux2 >> Java: AmazonCorretto11 >> Lucene/Solr: 8.8.2 >> >> We added the timeAllowed parameter to the request to Solr because we >> wanted to reduce the load on Solr. >> We set it to the same number as the request timeout to Solr. >> >> As a result, response times were significantly delayed, with the maximum >> value almost equal to the value specified for timeAllowed. >> The mean and median were degraded by less than 20%, the p99 were degraded >> by a factor of 1.5, the maximum value degraded by over a factor of 4. >> >> We removed the timeAllowed parameter from the request and the problem >> went away, so I guess timeAllowed is the cause of degradation. >> >> >> I am adding MINAMI who is a presenter, MIYAZAKI who is a co-worker, and >> my work email. >> Should you have any questions, please feel free to contact us. >> >> Best regards, >> Keisuke >> >> 2022年12月16日(金) 11:45 Koji Sekiguchi <[email protected]>: >> >>> Hi Alessandro, Mikhail, >>> >>> I thought that you were talking in the Tokyo meetup community mailing >>> list but now I realized that it was just emails :) >>> >>> I don't know the contact of the speaker, but I'm adding Isono-san who is >>> the leader of the search team in LIFULL. >>> >>> Hi Isono-san, >>> >>> They want to know the results of the investigation regarding >>> timeAllowed. And I want to know it, too. >>> >>> Koji >>> >>> 2022年12月16日(金) 10:16 Alessandro Benedetti <[email protected]>: >>> >>>> Hi Mikhail, Apache Solr Learning To Rank module *supports* distributed >>>> mode, it is the interleaving functionality that doesn't! >>>> >>>> As far as I know the Elasticsearch LTR works for multiple shards as >>>> well(the Open Source Connections plugin, as far as I know an official >>>> version is not out yet). >>>> >>>> In regards to the timeAllowed, @Koji Sekiguchi >>>> <[email protected]> can you add the presenter to this >>>> thread? >>>> >>>> Cheers >>>> -------------------------- >>>> *Alessandro Benedetti* >>>> Director @ Sease Ltd. >>>> *Apache Lucene/Solr Committer* >>>> *Apache Solr PMC Member* >>>> >>>> e-mail: [email protected] >>>> >>>> >>>> *Sease* - Information Retrieval Applied >>>> Consulting | Training | Open Source >>>> >>>> Website: Sease.io <http://sease.io/> >>>> LinkedIn <https://linkedin.com/company/sease-ltd> | Twitter >>>> <https://twitter.com/seaseltd> | Youtube >>>> <https://www.youtube.com/channel/UCDx86ZKLYNpI3gzMercM7BQ> | Github >>>> <https://github.com/seaseltd> >>>> >>>> >>>> On Thu, 15 Dec 2022 at 17:31, Mikhail Khludnev <[email protected]> wrote: >>>> >>>>> Not many replies, so far, though. >>>>> Alessandro, can you comment about your talk? I were surprised that >>>>> Solr LTR doesn't support distributed mode. Do you know about Elasticsearch >>>>> LTR? Does it work for multiple shards? >>>>> >>>>> >>>>> On Sun, Dec 11, 2022 at 6:28 AM Alessandro Benedetti < >>>>> [email protected]> wrote: >>>>> >>>>>> I second Mikhail, from the question I asked it seems the "time >>>>>> Allowed issue" that has occurred, it's currently under investigation, but >>>>>> it's clear to me either what's the performance issue discovered in the >>>>>> first place. >>>>>> Definitely interested in understanding it more, as some of our >>>>>> customers use the request param! >>>>>> Cheers >>>>>> >>>>>> On Sat, 10 Dec 2022, 15:22 Mikhail Khludnev, <[email protected]> wrote: >>>>>> >>>>>>> Hello, >>>>>>> Thanks for the notice. >>>>>>> At the second talk guys mentioned timeAllowed many times. I haven't >>>>>>> got the matter since I don't speak Japanese. Do they experience any >>>>>>> issues >>>>>>> with timeAllowed? I'm asking because I've fixed a few ill issues in it >>>>>>> some >>>>>>> time ago. >>>>>>> >>>>>>> >>>>>>> On Fri, Dec 9, 2022 at 3:18 AM Alessandro Benedetti < >>>>>>> [email protected]> wrote: >>>>>>> >>>>>>>> Hi guys, >>>>>>>> sorry for the spam but I wanted to let you know that tonight in >>>>>>>> Tokyo there's going to be the first Lucene/Solr in-person free meetup >>>>>>>> after >>>>>>>> a long time! >>>>>>>> >>>>>>>> https://lnkd.in/eDCXjdiX >>>>>>>> >>>>>>>> >>>>>>>> I am visiting Japan so it would be a pleasure to meet any member of >>>>>>>> the community that's around! >>>>>>>> >>>>>>>> This message is meant to be for anyone tonight in Tokyo, not >>>>>>>> necessarily committers/PMC members! >>>>>>>> >>>>>>>> >>>>>>>> Have a good one! >>>>>>>> -------------------------- >>>>>>>> *Alessandro Benedetti* >>>>>>>> Director @ Sease Ltd. >>>>>>>> *Apache Lucene/Solr Committer* >>>>>>>> *Apache Solr PMC Member* >>>>>>>> >>>>>>>> e-mail: [email protected] >>>>>>>> >>>>>>>> >>>>>>>> *Sease* - Information Retrieval Applied >>>>>>>> Consulting | Training | Open Source >>>>>>>> >>>>>>>> Website: Sease.io <http://sease.io/> >>>>>>>> LinkedIn <https://linkedin.com/company/sease-ltd> | Twitter >>>>>>>> <https://twitter.com/seaseltd> | Youtube >>>>>>>> <https://www.youtube.com/channel/UCDx86ZKLYNpI3gzMercM7BQ> | Github >>>>>>>> <https://github.com/seaseltd> >>>>>>>> >>>>>>> >>>>>>> >>>>>>> -- >>>>>>> Sincerely yours >>>>>>> Mikhail Khludnev >>>>>>> >>>>>> >>>>> >>>>> -- >>>>> Sincerely yours >>>>> Mikhail Khludnev >>>>> >>>> > > -- > Sincerely yours > Mikhail Khludnev > -- Sincerely yours Mikhail Khludnev
