Conversation
docs/ingestion/native-batch.md
Outdated
There was a problem hiding this comment.
The doc should address what happens if you disregard this advice and set appendToExisting: true but don't use dynamic partitioning. (Do you get an error? Does something weird happen?)
There was a problem hiding this comment.
I am still a noob, but I tried setting the partitionspec to hashed and appendToExisting to true. The task succeeded, but when I check the payload it looks like appendToExisting was set to false
"ioConfig": {
"type": "index_parallel",
"inputSource": {
"type": "http",
"uris": [
"https://static.imply.io/data/wikipedia.json.gz"
],
"httpAuthenticationUsername": null,
"httpAuthenticationPassword": null
},
"inputFormat": {
"type": "json",
"flattenSpec": {
"useFieldDiscovery": true,
"fields": []
},
"featureSpec": {}
},
"appendToExisting": false
I'll need to try again maybe and follow up with someone with deeper insight to verify how it is supposed to work in this case.
There was a problem hiding this comment.
@jihoonson PTAL. Clarified that the task will fail if you use another partitioning type and why. Included the error text in case someone needs to search on that. Thanks.
5845e59 to
9331331
Compare
docs/ingestion/native-batch.md
Outdated
There was a problem hiding this comment.
Actually there are a couple cases when something can go wrong.
appendToExisting: true,forceGuaranteedRollup: false,partitioning=dynamic: this is the only valid setup.appendToExisting: true,forceGuaranteedRollup: false,partitioning=hashed: error with a message ofDynamicPartitionsSpec must be used for best-effort rollup.appendToExisting: true,forceGuaranteedRollup: true,partitioning=dynamic: error with a message ofDynamicPartitionsSpec cannot be used for perfect rollup.appendToExisting: true,forceGuaranteedRollup: true,partitioning=hashed/single_dim: error with a message ofPerfect rollup cannot be guaranteed when appending to existing dataSources.
I'm not sure it's reasonable to list out all these errors here. It's hard to improve the error message for now because we don't know exactly what users want to do. Maybe it's better to simply say the task will fail.
BTW, forceGuaranteedRollup seems not very useful so we can get rid of it in the future. Once we remove it, then the error message can be simplified to something like appendToExisting cannot be set with hashed partitioning.
There was a problem hiding this comment.
thanks for the detailed review @jihoonson . I removed the line break and mention of error. I think your info can go into a troubleshooting topic on the forum
docs/ingestion/native-batch.md
Outdated
There was a problem hiding this comment.
Did you intend to break the line here? This actually breaks the table (https://github.com/apache/druid/blob/9331331b857419c6ca45c15a1852058355019c87/docs/ingestion/native-batch.md#ioconfig). You need to add <br/> to break the line.
9331331 to
421ae55
Compare
jihoonson
left a comment
There was a problem hiding this comment.
The new doc looks good to me, but please apply the same change to the other place as well. Also please fix the typo reported in the CI.
721 | and log the following error: "forceGuranteedRollup must be set". |false|no|
docs/ingestion/native-batch.md
Outdated
There was a problem hiding this comment.
I think you want to apply the same change here as well?
421ae55 to
4834d28
Compare
|
The doc CI passed. I'm merging this PR. |
Fixes awkward wording around the behavior of
appendToExsiting.Description
Removes ambiguity around the "current limitation". Describes the option in terms of the current product functionaltiy.
This PR has: