Dfs query phase coordinator metric #136481
Conversation
elasticsearchmachine
added
v9.3.0
needs:triage
chrisparrinello
added
>enhancement
Team:Search Foundations
|
Pinging @elastic/es-search-foundations (Team:Search Foundations) |
|
Hi @chrisparrinello, I've created a changelog YAML for you. |
| private final Client client; | ||
| private final AbstractSearchAsyncAction<?> context; | ||
| private final SearchProgressListener progressListener; | ||
| private long phaseStartTimeInNanos; |
There was a problem hiding this comment.
where do you set the initial value for this? I guess it's missing and tests don't catch it because we never verify that timing is sensible :)
There was a problem hiding this comment.
Whoops. Yeah it was missed. I'll add it.
| } | ||
|
|
||
| private void onFinish(AggregatedDfs dfs) { | ||
| context.getSearchResponseMetrics().recordSearchPhaseDuration(getName(), System.nanoTime() - phaseStartTimeInNanos); |
There was a problem hiding this comment.
For this one, shouldn't we be able to store attributes like we do at the shard level? I may have forgotten to raise this in your previous PR. perhaps a follow-up for later to do so when possible?
There was a problem hiding this comment.
I think that would be best for a follow-up PR for later. I think the issue in some cases is having access to the SearchRequest and the list of localIndices. I think the SearchRequest object is easy to come by but the localIndices not as easy. If it is okay to pass in an empty array or null then we can take care of this sooner I think.
3 participants
The query phase of a DFS search runs in a separate implementation. This change adds a new metric
es.search_response.took_durations.dfs_query.histogramto measure that.