• 2
    • 27
    • We collect Jira feedback from various sources, and we evaluate what we've collected when planning our product roadmap. To understand how this piece of feedback will be reviewed, see our Implementation of New Features Policy.

      Problem Definition

      Currently when the number of characters on the lengthiest rank field on the AO_60DB71_LEXORANK table exceeds:

      That will lead to the following:

      • Lexorank rebalance job will balance whole space for RANK and that means touching all issues.
      • For large instance this might take significant time (days).

      In some cases, those long RANK values are localised in small area and there is no need to rebalance whole space.
      For example, there is a problem only around space hz :

      1|zzzzzz: 
      1|w00000:                     
      1|hzzzzz:                             
      1|hzzzzy:zzzzzzzzzzzzzzzzzzzzzzzzzzzzy
      1|hzzzzy:zzzzzzzzzzzzzzzzzzzzzzzzzzzzw
      1|hzzzzy:zzzzzzzzzzzzzzzzzzzzzzzzzzzzr
      1|i00000:
      

      In majority cases there are problem with a couple of thousands rows only, while whole table has millions rows.

      Suggested Solution

      Add option for Lexorank rebalancing task for localize those areas and rebalance only them.
      That will significantly reduce length of rebalancing (from days to minutes). As a consequence it reduce load for Jira.

      Workaround

      None

            [JSWSERVER-16471] Add option to do local Lexorank rebalancing

            Gaven Ray added a comment -

            We are facing the issue currently on 7.8.0 where users can't Rank and Jira claims it is rebalancing. 

            Database Statistics

            Attachments 95439
            Comments 730381
            Components 1330
            Custom Fields 1227
            Groups 644
            Issue Security Levels 9
            Issue Types 466
            Issues 432001
            Priorities 8
            Projects 341
            Resolutions 17
            Screen Schemes 338
            Screens 487
            Statuses 252
            Users 4765
            Versions 4137
            Workflows 318

            I am aware we need to cut down on the custom fields. We are testing some fields to do that. It would be great if Rank was not based on every project in Jira but on a project by project level.

            Gaven Ray added a comment - We are facing the issue currently on 7.8.0 where users can't Rank and Jira claims it is rebalancing.  Database Statistics Attachments 95439 Comments 730381 Components 1330 Custom Fields 1227 Groups 644 Issue Security Levels 9 Issue Types 466 Issues 432001 Priorities 8 Projects 341 Resolutions 17 Screen Schemes 338 Screens 487 Statuses 252 Users 4765 Versions 4137 Workflows 318 I am aware we need to cut down on the custom fields. We are testing some fields to do that. It would be great if Rank was not based on every project in Jira but on a project by project level.

            Dom Baldin [Adobe] added a comment - - edited

            This would be extremely helpful. Our DC with 4.9M issues takes around 4 weeks to complete. Our customers definitely feel it and the platform is more unstable overall. We've had to suspend Lexo.

            We are running 7.6.4 currently and are running into an issue where Lexo suspends frequently due to index replication delays. It spends more time suspended then it does balancing. I suspect the index replication is happening due to the size of our index. 

            Certainly, this feature would help us alleviate this.

            Dom Baldin [Adobe] added a comment - - edited This would be extremely helpful. Our DC with 4.9M issues takes around 4 weeks to complete. Our customers definitely feel it and the platform is more unstable overall. We've had to suspend Lexo. We are running 7.6.4 currently and are running into an issue where Lexo suspends frequently due to index replication delays. It spends more time suspended then it does balancing. I suspect the index replication is happening due to the size of our index.  Certainly, this feature would help us alleviate this.

            Very useful. Rebalancing for us takes between 8 and 10 days.

            Greg Warner (Amazon) added a comment - Very useful. Rebalancing for us takes between 8 and 10 days.

            Matt Doar added a comment -

            This be very helpful in large Jira instances. DC with 4M issues takes days and days for a Lexorank rebalance job

            Matt Doar added a comment - This be very helpful in large Jira instances. DC with 4M issues takes days and days for a Lexorank rebalance job

            I am very happy to see this suggestion.
            This has been a major pain point for our team as the re balance takes multiple days almost every month.

            Alex Johnson added a comment - I am very happy to see this suggestion. This has been a major pain point for our team as the re balance takes multiple days almost every month.

              Unassigned Unassigned
              ayakovlev@atlassian.com Andriy Yakovlev [Atlassian]
              Votes:
              35 Vote for this issue
              Watchers:
              31 Start watching this issue

                Created:
                Updated: