capistrant opened a new pull request #10284: URL: https://github.com/apache/druid/pull/10284
<!-- Thanks for trying to help us make Apache Druid be the best it can be! Please fill out as much of the following information as is possible (where relevant, and remove it when irrelevant) to help make the intention and scope of this PR clear in order to ease review. --> <!-- Replace XXXX with the id of the issue fixed in this PR. Remove this section if there is no corresponding issue. Don't reference the issue in the title of this pull-request. --> <!-- If you are a committer, follow the PR action item checklist for committers: https://github.com/apache/druid/blob/master/dev/committer-instructions.md#pr-and-issue-action-item-checklist-for-committers. --> ### Description <!-- Describe the goal of this PR, what problem are you fixing. If there is a corresponding issue (referenced above), it's not necessary to repeat the description here, however, you may choose to keep one summary sentence. --> A large cluster with many segments results in a lot of work being done by the Coordinator in order to complete its duties. I believe that any optimization to coordinator duties can help in a large cluster. This patch gives an experienced admin a knob to turn in order to try and shave some time off of the balance segments duty. As of now, existing Balancer Strategies iterate over all of the segments in the cluster when choosing a segment to move. The first segment candidate is the most likely to be moved and the last segment candidate is the least likely to move. This patch gives an admin the ability to put a limit on the number of segments that will be candidates to be moved. For most cases, I don't think this knob will be needed, but in some large enterprise cases I feel that it could be beneficial. <!-- Describe your patch: what did you change in code? How did you fix the problem? --> I updated the BalancerStrategy Interface. The pickSegmentToMove method gained a 3rd parameter that specifies the number of segments that should be considered when picking a segment to move. `CostBalancerStrategy` (and it's inheriting classes) and `RandomBalancerStrategy` both leverage `ReservoirSegmentSampler` to choose a segment "at random" from a list of candidate servers. I updated the required method in `ReservoirSegmentSampler` to adhere to the limiting parameter described above. If the limit is reached, the method picking a segment will break out of its iteration and return immediately. Currently all code paths use a new dynamic coordinator config that an admin can tune if they'd like to put a limiter on this action of picking a segment to move. The default value for the config is such that all segments will be iterated and be candidates to pick. I thought a dynamic config was good because it is flexible and could be leveraged in times such as if you wanted to temporarily boost up the number of segments to move in order rebalance to new servers faster. If doing that, and you also wanted to make this go quicker by not bothering with having so many potential segments to be picked to move. The new dynamic config is `maxSegmentsToConsiderPerMove` with a default of `Integer.MAX_VALUE` I call out in the documentation that an admin should be experienced when considering altering this config. I say that because in many cases, the default is fine. <!-- In each section, please describe design decisions made, including: - Choice of algorithms - Behavioral aspects. What configuration values are acceptable? How are corner cases and error conditions handled, such as when there are insufficient resources? - Class organization and design (how the logic is split between classes, inheritance, composition, design patterns) - Method organization and design (how the logic is split between methods, parameters and return types) - Naming (class, method, API, configuration, HTTP endpoint, names of emitted metrics) --> <!-- It's good to describe an alternative design (or mention an alternative name) for every design (or naming) decision point and compare the alternatives with the designs that you've implemented (or the names you've chosen) to highlight the advantages of the chosen designs and names. --> An alternative of this approach would be to restrict what is sent to `pickSegmentToMove` in the first place. I choose not to approach this at this time because I didn't like the idea of either choosing the number of `ServerHolders` to send to `pickSegmentsToMove` or to analyze the `ServerHolders` before picking how many to send to ensure only a certain number of segments are sent. I'd be open to re-assessing whether or not this would be a better approach or not if someone suggests it may be the proper approach. <!-- If there was a discussion of the design of the feature implemented in this PR elsewhere (e. g. a "Proposal" issue, any other issue, or a thread in the development mailing list), link to that discussion from this PR description and explain what have changed in your final design compared to your original proposal or the consensus version in the end of the discussion. If something hasn't changed since the original discussion, you can omit a detailed discussion of those aspects of the design here, perhaps apart from brief mentioning for the sake of readability of this PR description. --> <!-- Some of the aspects mentioned above may be omitted for simple and small changes. --> <hr> This PR has: - [X] been self-reviewed. - [X] added documentation for new or modified features or behaviors. - [X] added Javadocs for most classes and all non-trivial methods. Linked related entities via Javadoc links. - [X] added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar reader. - [X] added unit tests or modified existing tests to cover new code paths, ensuring the threshold for [code coverage](https://github.com/apache/druid/blob/master/dev/code-review/code-coverage.md) is met. - [X] been tested in a test Druid cluster. <!-- Check the items by putting "x" in the brackets for the done things. Not all of these items apply to every PR. Remove the items which are not done or not relevant to the PR. None of the items from the checklist above are strictly necessary, but it would be very helpful if you at least self-review the PR. --> <hr> ##### Key changed/added classes in this PR * `BalancerStrategy` Interface * `CostBalancerStrategy` * `RandomBalancerStrategy` * `ReservoirSegmentSampler` * `CoordinatorDynamicConfig` ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
