You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@druid.apache.org by GitBox <gi...@apache.org> on 2021/04/19 23:10:53 UTC
[GitHub] [druid] capistrant opened a new pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
capistrant opened a new pull request #11135:
URL: https://github.com/apache/druid/pull/11135
<!-- 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. -->
<!-- Please read the doc for contribution (https://github.com/apache/druid/blob/master/CONTRIBUTING.md) before making this PR. Also, once you open a PR, please _avoid using force pushes and rebasing_ since these make it difficult for reviewers to see what you've changed in response to their reviews. See [the 'If your pull request shows conflicts with master' section](https://github.com/apache/druid/blob/master/CONTRIBUTING.md#if-your-pull-request-shows-conflicts-with-master) for more details. -->
<!-- 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. -->
<!-- Describe your patch: what did you change in code? How did you fix the problem? -->
<!-- If there are several relatively logically separate changes in this PR, create a mini-section for each of them. For example: -->
Add a new dynamic configuration to the coordinator that gives an operator the power to set a hard limit for the number of non-primary segment replicas that are loaded during a single execution of `RunRules#run`. This allows the operator to limit the amount of work loading non-primary replicas that `RunRules` will execute in a single run. An example of a reason to use a non-default value for this new config is if the operator wants to ensure that major events such as historical service(s) leaving the cluster, large ingestion jobs, etc. do not cause an abnormally long `RunRules` execution compared to the cluster's baseline runtime.
**Example**
cluster: 3 historical servers in _default_tier with 18k segments per server. Each segment belongs to a datasource that has the load rule "LoadForever 2 replicas on _default_tier". The cluster load status is 100% loaded.
Event: 1 historical drops out of the cluster.
Today: The coordinator will load all 18k segments that are now under-replicated in a single execution of RunRules (as long as Throttling limits are not hit and there is capacity)
My change: The coordinator can load a limited number of these under-replicated segments IF the operator has tuned the new dynamic config down from its default. For instance, the operator could say that it is 2k. Meaning it would take at least 9 coordination cycles to fully replicate the segments that were on the recently downed host.
**Why**
Operators need to balance lots of competing needs. Having the cluster fully replicated is great for HA. But if an event causes the coordinator to take 20 minutes to fully replicate because it has to load thousands of replicas, we sacrifice the timeliness of loading newly ingested segments that were inserted into the metastore after this long coordination cycle started. Maybe the operator cares more about that fresh data timeliness than the replication status, so they change the new config to a value that causes RunRules to take less time but require more execution cycles to bring the data back to full replication.
Really what the change aims to do is give an operator more flexibility. As written the default would give the operator the exact same functionality that they see today.
<!--
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)
-->
**Design**
I folded this new configuration and feature into ReplicationThrottler. That is essentially what it is doing, just in a new way compared to the current ReplicationThrottler functionality.
<!-- 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. -->
<!-- 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>
##### Key changed/added classes in this PR
* `CoordinatorDynamicConfig`
* `ReplicationThrottler`
* `RunRules`
* `LoadRule`
<hr>
<!-- 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 below are strictly necessary, but it would be very helpful if you at least self-review the PR. -->
This PR has:
- [ ] been self-reviewed.
- [ ] using the [concurrency checklist](https://github.com/apache/druid/blob/master/dev/code-review/concurrency.md) (Remove this item if the PR doesn't have any relation to concurrency.)
- [ ] added documentation for new or modified features or behaviors.
- [ ] added Javadocs for most classes and all non-trivial methods. Linked related entities via Javadoc links.
- [ ] added or updated version, license, or notice information in [licenses.yaml](https://github.com/apache/druid/blob/master/dev/license.md)
- [ ] added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar reader.
- [ ] 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.
- [ ] added integration tests.
- [ ] been tested in a test Druid cluster.
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] capistrant merged pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
capistrant merged pull request #11135:
URL: https://github.com/apache/druid/pull/11135
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] capistrant commented on a change in pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
capistrant commented on a change in pull request #11135:
URL: https://github.com/apache/druid/pull/11135#discussion_r620478702
##########
File path: server/src/main/java/org/apache/druid/server/coordinator/CoordinatorDynamicConfig.java
##########
@@ -176,6 +184,22 @@ public CoordinatorDynamicConfig(
}
this.pauseCoordination = pauseCoordination;
this.replicateAfterLoadTimeout = replicateAfterLoadTimeout;
+
+ if (maxNonPrimaryReplicantsToLoad == null) {
Review comment:
Hmm, I think I would be ok with that. I don't see any valid use case where a value of 0 would be required by the user. At that point they would want to disable replication via load rules.
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] a2l007 commented on a change in pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
a2l007 commented on a change in pull request #11135:
URL: https://github.com/apache/druid/pull/11135#discussion_r620474349
##########
File path: server/src/main/java/org/apache/druid/server/coordinator/CoordinatorDynamicConfig.java
##########
@@ -176,6 +184,22 @@ public CoordinatorDynamicConfig(
}
this.pauseCoordination = pauseCoordination;
this.replicateAfterLoadTimeout = replicateAfterLoadTimeout;
+
+ if (maxNonPrimaryReplicantsToLoad == null) {
Review comment:
Should we consider using 0 as an non-configured value and change the check here? That would avoid the primitive type change.
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] a2l007 commented on pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
a2l007 commented on pull request #11135:
URL: https://github.com/apache/druid/pull/11135#issuecomment-823420836
Thanks for the PR! This config should come in handy to reduce coordinator churn in case historicals fall out of the cluster. Have you thought about configuring `maxNonPrimaryReplicantsToLoad` specific to a tier instead of a global property?
Also could you please add some docs related to this property to the configuration docs?
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] capistrant commented on pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
capistrant commented on pull request #11135:
URL: https://github.com/apache/druid/pull/11135#issuecomment-823553928
> Thanks for the PR! This config should come in handy to reduce coordinator churn in case historicals fall out of the cluster. Have you thought about configuring `maxNonPrimaryReplicantsToLoad` specific to a tier instead of a global property?
> Also could you please add some docs related to this property to the configuration docs?
I added the missing docs.
I had not thought about making this a per-tier setting. I'm coming at it from the angle of an operator not caring if the non-primary replicants are in tier X, Y, or Z, but rather just wanting to make sure the coordinator never spends too much time loading these segments and not doing its other jobs, mainly discovering and loading newly ingested segments.
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] a2l007 commented on a change in pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
a2l007 commented on a change in pull request #11135:
URL: https://github.com/apache/druid/pull/11135#discussion_r617042227
##########
File path: docs/configuration/index.md
##########
@@ -822,6 +823,7 @@ Issuing a GET request at the same URL will return the spec that is currently in
|`decommissioningMaxPercentOfMaxSegmentsToMove`| The maximum number of segments that may be moved away from 'decommissioning' servers to non-decommissioning (that is, active) servers during one Coordinator run. This value is relative to the total maximum segment movements allowed during one run which is determined by `maxSegmentsToMove`. If `decommissioningMaxPercentOfMaxSegmentsToMove` is 0, segments will neither be moved from _or to_ 'decommissioning' servers, effectively putting them in a sort of "maintenance" mode that will not participate in balancing or assignment by load rules. Decommissioning can also become stalled if there are no available active servers to place the segments. By leveraging the maximum percent of decommissioning segment movements, an operator can prevent active servers from overload by prioritizing balancing, or decrease decommissioning time instead. The value should be between 0 and 100.|70|
|`pauseCoordination`| Boolean flag for whether or not the coordinator should execute its various duties of coordinating the cluster. Setting this to true essentially pauses all coordination work while allowing the API to remain up. Duties that are paused include all classes that implement the `CoordinatorDuty` Interface. Such duties include: Segment balancing, Segment compaction, Emission of metrics controlled by the dynamic coordinator config `emitBalancingStats`, Submitting kill tasks for unused segments (if enabled), Logging of used segments in the cluster, Marking of newly unused or overshadowed segments, Matching and execution of load/drop rules for used segments, Unloading segments that are no longer marked as used from Historical servers. An example of when an admin may want to pause coordination would be if they are doing deep storage maintenance on HDFS Name Nodes with downtime and don't want the coordinator to be directing Historical Nodes to hit the Name Node with API req
uests until maintenance is done and the deep store is declared healthy for use again. |false|
|`replicateAfterLoadTimeout`| Boolean flag for whether or not additional replication is needed for segments that have failed to load due to the expiry of `druid.coordinator.load.timeout`. If this is set to true, the coordinator will attempt to replicate the failed segment on a different historical server. This helps improve the segment availability if there are a few slow historicals in the cluster. However, the slow historical may still load the segment later and the coordinator may issue drop requests if the segment is over-replicated.|false|
+|`maxNonPrimaryReplicantsToLoad`|This is the maximum number of non-primary segment replicants to load per Coordination run. This number can be set to put a hard upper limit on the number of replicants loaded. It is a tool that can help prevent long delays in new data being available for query after events that require many non-primary replicants to be loaded by the cluster; such as a Historical node disconnecting from the cluster. The default value essentially means there is no limit on the number of replicants loaded per coordination cycle.|`Integer.MAX_VALUE`|
Review comment:
It would be useful if we add some info regarding what could be a good starting value to set this to.
##########
File path: server/src/main/java/org/apache/druid/server/coordinator/duty/RunRules.java
##########
@@ -128,6 +130,18 @@ public DruidCoordinatorRuntimeParams run(DruidCoordinatorRuntimeParams params)
boolean foundMatchingRule = false;
for (Rule rule : rules) {
if (rule.appliesTo(segment, now)) {
+ if (
+ stats.getGlobalStat(
+ "totalNonPrimaryReplicantsLoaded") >= paramsWithReplicationManager.getCoordinatorDynamicConfig()
+ .getMaxNonPrimaryReplicantsToLoad()
+ && !paramsWithReplicationManager.getReplicationManager().isLoadPrimaryReplicantsOnly()
+ ) {
+ log.info(
+ "Maximum number of non-primary replicants [%d] have been loaded for the current RunRules execution. Only loading primary replicants from here on.",
Review comment:
Since this behavior is valid only for the present coordinator run, the log message might be clearer with something like "Only loading primary replicants from here on for this coordinator run period"
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] a2l007 commented on pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
a2l007 commented on pull request #11135:
URL: https://github.com/apache/druid/pull/11135#issuecomment-831942053
@capistrant Yup, LGTM. Thanks!
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] capistrant commented on pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
capistrant commented on pull request #11135:
URL: https://github.com/apache/druid/pull/11135#issuecomment-831907264
@a2l007 are you okay with merge this week now that the issue for pursuing a cleaner configuration strategy is created?
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] capistrant commented on a change in pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
capistrant commented on a change in pull request #11135:
URL: https://github.com/apache/druid/pull/11135#discussion_r620485631
##########
File path: server/src/main/java/org/apache/druid/server/coordinator/CoordinatorDynamicConfig.java
##########
@@ -176,6 +184,22 @@ public CoordinatorDynamicConfig(
}
this.pauseCoordination = pauseCoordination;
this.replicateAfterLoadTimeout = replicateAfterLoadTimeout;
+
+ if (maxNonPrimaryReplicantsToLoad == null) {
Review comment:
although doing this would kind of hide a user error. If they submit 0 but we change 0 to the default and log it, they wouldn't know 0 is invalid.
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] capistrant commented on pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
capistrant commented on pull request #11135:
URL: https://github.com/apache/druid/pull/11135#issuecomment-826945930
https://github.com/apache/druid/blob/master/server/src/main/java/org/apache/druid/server/coordinator/CoordinatorDynamicConfig.java#L141
This PR has a similar issue that resulted in this block of code. I think I will do the same solution for now. but long term it would be cool if this had a more elegant solution.
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
[GitHub] [druid] a2l007 commented on a change in pull request #11135: Create dynamic config that can limit number of non-primary replicants loaded per coordination cycle
Posted by GitBox <gi...@apache.org>.
a2l007 commented on a change in pull request #11135:
URL: https://github.com/apache/druid/pull/11135#discussion_r620535314
##########
File path: server/src/main/java/org/apache/druid/server/coordinator/CoordinatorDynamicConfig.java
##########
@@ -176,6 +184,22 @@ public CoordinatorDynamicConfig(
}
this.pauseCoordination = pauseCoordination;
this.replicateAfterLoadTimeout = replicateAfterLoadTimeout;
+
+ if (maxNonPrimaryReplicantsToLoad == null) {
Review comment:
Yeah I'm fine with leaving it as `Integer` until we have a better solution in place to fix the dynamic config behavior during upgrade. It would be useful to log an issue for that behavior in case somebody would like to work on it.
--
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:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org