I suppose the biggest spots for peer review are: * use of brackets [ ] in the metric name where the request handler is. Thus "/select[shard]" * There is a fundamental difference in how the metrics are tracked on a handler. Previously, there were metrics for all of /select (no matter how it was invoked), and a few for .distrib. & .shard. depending on how it was invoked. Now, the request is classified to be a shard request, or not a shard request, after which separate metrics (same type/semantics) are manipulated based on that classification, kind of as if there are two distinct request handlers even though just one is registered. I think the PR makes this clear. While I like it, the main trade-off is that a user would be forced to aggregate metrics if they wanted a single metric for the handler. I think the isShard=true request changes the personality/mode of the handler so much that I prefer to present it as its own identity from a metrics standpoint.
~ David Smiley Apache Lucene/Solr Search Developer http://www.linkedin.com/in/davidwsmiley On Tue, Mar 1, 2022 at 4:19 PM Timothy Potter <thelabd...@gmail.com> wrote: > Hi David, > > I read your note about SOLR-14401 but not clear what you need from me? > Seems like you're renaming existing metrics and removing "distrib" > from handlers that don't support a distrib mode, seems right to me. > > I actually haven't done much work on the metrics backend. For Grafana, > it's a JSON file so search / replace the metrics you're changing. The > Solr operator makes it really easy to set up Solr + ZK + Grafana + > Prometheus + Exporter to test out your changes. It'll be pretty > obvious if the dashboard is broken. > > Tim > > On Tue, Mar 1, 2022 at 7:01 AM David Smiley <dsmi...@apache.org> wrote: > > > > > > > > On Tue, Mar 1, 2022 at 4:46 AM Jan Høydahl <jan....@cominvent.com> > wrote: > >> > >> Hi, and welcome to March! > >> > >> Our initial goal of a RC1 within February slipped, but we are still in > a good position. > >> I'll try to summarize the current code blockers: > >> > >> > >> SOLR-16061 Decouple CloudSolrClient from ZkStateReader > >> > >> This is new, a spin-off from SOLR-15342 to prepare for solrj > modularization. There is already a draft PR. Hope there will be progress on > this so we don't have to delay solrj modularization until 10.0 > > > > > > I'm working with Haythem on this (a colleague). I think it's close; > it's "just" a refactoring. The main constraint on this is Haythem's time. > > > >> > >> > >> SOLR-14290 Fix NPE in SolrTestCaseJ4 breaking external usage for > master/9.x > >> > >> This has not seen any movement despite repeated reminders, so unless > there is progress within a few days I'll remove it as blocker and add a > note to the release notes that users relying on running test framework > locally should wait for a later release. > > > > > > I'm interested in looking but not until I get through the other two. > > > >> > >> SOLR-14401 "distrib" request handler metrics should only be tracked > on... > >> > >> There is a PR, not sure how close to merge it is though. > > > > > > I think the core of the change is fine but there are downstream changes > needed. First level is the prometheus exporter configuration to not look > for ".distrib." vs ".local."; it's different now. Next level is the > Grafana dashboard. I don't normally play with JQ, Prometheus or Grafana so > it's taking me some time this week. I'd appreciate any feedback on the > choices here; so far only Houston has weigh'ed in. I think Tim Potter, if > you're reading this, would be useful given you did major work here. > > > >> > >> > >> Also, David found a new blocker bug yesterday - the "jetty" metrics > group is missing in 9.x. There will likely be another blocker due to this. > >> > >> > >> Appreciate an update in this thread on the ETA for each of these. > >> > >> Jan > >> > >> 22. feb. 2022 kl. 12:20 skrev Jan Høydahl <jan....@cominvent.com>: > >> > >> I created a new blocker > >> > >> SOLR-16040 Fix split packages in hdfs module > >> > >> Not sure if it needs to be a blocker though, but we should try to avoid > split packages as far as we can, and this cannot be done in 9.x. > >> Meanwhile, SOLR-15064 is resolved and SOLR-14401 is in PR review phase. > >> Jenkins is now mostly green after some turmoil! > >> > >> SOLR-14290 (SolrTestCaseJ4 NPE) seems to be stalled - anyone who can > lend a hand there? > >> > >> We also discussed in SOLR-15342 whether refactoring CloudSolrClient to > untangle ZkStateProvider should be done now, and also rename solr-solrj as > solr-solrj-all so that we can continue with the solrj modularization in 9.x > without back-compat breaks. > >> It seems worthy of a blocker to me, but we need someone willing to do > the work in the next few days. Anyone? > >> > >> I also created SOLR-16041 (not blocker) to try to setup nightly > smoketestRelease Jenkins jobs, I may try to give it a go. > >> > >> Assuming progress on the above, I'm still hopeful for an RC1 in the > timeframe of next week. > >> > >> Jan > >> > >> 16. feb. 2022 kl. 17:05 skrev Jan Høydahl <jan....@cominvent.com>: > >> > >> These are the three main code-blockers for doing 9.0.0 RC1: > >> > >> (P) SOLR-15064 Atomic/partial updates to nested docs should not assume > _route_ dsmiley dsmiley > >> (S) SOLR-14290 Fix NPE in SolrTestCaseJ4 breaking external usage for > master/9.x gus gus > >> (S) SOLR-14401 """distrib"" request handler metrics should only be > tracked on pe dsmiley dsmiley > >> > >> The other blockers are mostly about the release process itself, > including docker and refguide. I'm doing a clean-up of 9.0 CHANGES too. > >> When these are resolved, I'll prepare RC1. That means we're really > close now!! > >> > >> Anshum is preparing a release notes draft, and we also need to complete > "Major Changes in 9.0" and "Upgrade Notes" in ref-guide before publishing > the guide. > >> > >> Jan > >> > >> 7. feb. 2022 kl. 14:52 skrev Jan Høydahl <jan....@cominvent.com>: > >> > >> Congrats on HDFS as a package! Huge win! Also some other blockers have > been closed recently. > >> > >> Status on the 9.0 release, one week into February. > >> > >> - I have done a dry-run of an RC and the smoketester. Think the release > scripts are ready! > >> - 11 open blockers: > >> > >> (P) SOLR-15587 Replicas end up with base_url as http on client side > even if clus thelabdude thelabdude > >> (P) SOLR-15556 Ref Guide Redesign Phase 3: Replace Jekyll > ctargett ctargett > >> (P) SOLR-15557 Figure out how to handle ref guide page > renames/redirects janhoy ctargett > >> (A) SOLR-15064 Atomic/partial updates to nested docs should not assume > _route_ dsmiley dsmiley > >> (A) SOLR-15949 Use Java 17 in docker > dsmiley dsmiley > >> (S) SOLR-14290 Fix NPE in SolrTestCaseJ4 breaking external usage for > master/9.x gus gus > >> (S) SOLR-14401 """distrib"" request handler metrics should only be > tracked on pe dsmiley dsmiley > >> (S) SOLR-15321 "Flesh out process for managing/storing ""official"" > Dockerfiles houston hossman > >> (U) SOLR-15941 Wire Antora refGuide into the release scripts > janhoy > >> (U) SOLR-15898 Complete Major changes and Upgrade Notes in RefGudie > for 9.0.0 janhoy > >> (D) SOLR-13138 Remove deprecated code prior to 9.0 > romseygeek > >> > >> Based on a quick look, I assigned a readiness-code for each: > >> (A): Active, pending PR (P): PR in progress, (D): Done, can be closed? > (S): Stalled. (U): Unassigned > >> > >> Would be great to have a status and ETA for the (S)talled issues. > Perhaps we can push some to 9.0.1 or 9.1? > >> > >> Wrt publishing the ref-guide and docker image, those could be done > semi-manual for 9.0.0 if need be. > >> > >> Jan > >> > >> 31. jan. 2022 kl. 21:54 skrev Kevin Risden <compuwizard...@gmail.com>: > >> > >>> SOLR-14660 Migrating HDFS into a module > krisden ichattopadhyaya > >> > >> > >> This was just merged. Big shout out to Istvan Farkas for doing 99% of > the heavy lifting here! > >> > >> Kevin Risden > >> > >> > >> On Sat, Jan 29, 2022 at 5:11 PM Jan Høydahl <jan....@cominvent.com> > wrote: > >>> > >>> Hi, > >>> > >>> Nine days ago we had 24 blockers. Several blockers have since been > added and several resolved, and today we are at 19. > >>> > >>> I notified in the previous email that I'd remove inactive blockers. > Here is the list of JIRAs for which I intend to remove the blocker flag on > February 1st: > >>> > >>> SOLR-14097 coreRootDirectory should be solr_home/cores > >>> SOLR-15064 Atomic/partial updates to nested docs should not assume > _route_ param is the root ID > >>> SOLR-15242 Consolidate README.md with solr/README.md > >>> SOLR-15096 [REGRESSION] Collection Delete Performance significantly > degraded in Java 11 v 8 > >>> SOLR-15223 Deprecate HttpSolrClient, mark httpcomponents dep as > "optional" in SolrJ > >>> SOLR-15835 Collection creation failing with https > >>> > >>> Shout out if you are working on one of these and expect it to be > finished soon-ish. > >>> > >>> That leaves 13 blockers: > >>> > >>> Issue key Summary > Assignee Reporter > >>> SOLR-14660 Migrating HDFS into a module > krisden ichattopadhyaya > >>> SOLR-15956 Add documentation for creating a docker image from the > binary dis houston houston > >>> SOLR-13138 Remove deprecated code prior to 9.0 > romseygeek > >>> SOLR-15556 Ref Guide Redesign Phase 3: Replace Jekyll > ctargett ctargett > >>> SOLR-15949 Use Java 17 in docker > dsmiley dsmiley > >>> SOLR-15926 Fix version specification in the Solr Ref Guide > houston > >>> SOLR-14290 Fix NPE in SolrTestCaseJ4 breaking external usage for > master/9.x gus gus > >>> SOLR-12901 Make UnifiedHighlighter the default > dsmiley dsmiley > >>> SOLR-14401 """distrib"" request handler metrics should only be > tracked on pe dsmiley dsmiley > >>> SOLR-15587 Replicas end up with base_url as http on client side even > if clus thelabdude thelabdude > >>> SOLR-15557 Figure out how to handle ref guide page renames/redirects > ctargett > >>> SOLR-15898 Complete Major changes and Upgrade Notes in RefGudie for > 9.0.0 janhoy > >>> SOLR-15321 "Flesh out process for managing/storing ""official"" > Dockerfiles houston hossman > >>> > >>> It seems like most of these have had some recent activity. > >>> Please everyone, have a look if you can lend a hand with any of these, > so we can get the list to zero early in February and do the first RC. > >>> I'd appreciate some help on SOLR-15898, consolidating and structuring > the "Major changes" chapter of the reference guide. > >>> > >>> Jan > >>> > >>> > >>> 20. jan. 2022 kl. 19:48 skrev Jan Høydahl <jan....@cominvent.com>: > >>> > >>> Hi, > >>> > >>> The list of release blockers can be seen with this JIRA filter: > https://issues.apache.org/jira/issues/?filter=12351219 > >>> > >>> After the committer's meeting today we decided to add these to the > blockers list: > >>> - SOLR-15556 Ref Guide Redesign Phase 3: Replace Jekyll > >>> - SOLR-15917 Rename 'contrib' as 'module' > >>> - SOLR-15880 Introduce Support to K Nearest Neighbors Search > >>> - SOLR-14660 Migrating HDFS into a package > >>> - SOLR-12901 Make UnifiedHighlighter the default > >>> - SOLR-15914 Make it super simple to add a contrib module to shared > classpath > >>> - And probably some minor ones too > >>> > >>> Each of these are in-flight and are expected to be ready really soon™. > >>> > >>> There are currently 24 blockers, but some of those are Unsassigned > and/or have not been given any attention for some time. > >>> On Feb 1st I'll take the freedom to remove blocker flag for those that > have not moved anywhere since. > >>> If you want to own one of them, please assign yourself and communicate > progress and an ETA. > >>> > >>> Jan > >>> > >>> > >> > >> > >> > >> > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@solr.apache.org > For additional commands, e-mail: dev-h...@solr.apache.org > >
