You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
LogAttribute parameters autocomplete is not working #2163
It looks like autocomplete for attribute names is not working right.
Some fields in LogAttributes will never get suggested for search and will never appear in the available filters even when the number of filters to fetch is high (its 1000 in the example).
Example: A log has the attribute "organization" in its LogAttribute structure.
The autocomplete is not suggesting the "organization" attribute.
The filters for LogAttribute show a very small subset of all the LogAttribute attributes even when the amount of filters to be fetched is 1000.
The "organization" attribute is not there and also all the other _usage params are not there.
That PR will get us part of the way there, I have a follow on planned that will enable those MVs to be used also for filters.
Introspection queries are really intensive, it's actually a quite difficult problem. MVs should trivialize the problem of fetching all values as part of a set. Subsets may still be difficult, but we'll introduce a toggle that allows you to still surface the "exact" values.
Hey @czephyr, wanted to let you know that there are some MVs you could add today to assist. It'll require you running some SQL on your CH instance and modifying your source in your team settings.
The file can be found here. #2128 is where the MVs were added, there's a video showing how you can add to the source.
I've added the MVs to my deployment making them ReplicatedSummingMergeTree and pointing the HDX metadata tables to the distributed version of them.
I'm running into this error:
Code: 70. DB::Exception: Received from localhost:9000. DB::Exception: Conversion from AggregateFunction(groupUniqArray(20), String) to AggregateFunction(groupUniqArray(20), String) is not supported: While executing Remote. (CANNOT_CONVERT_TYPE)
It looks like HDX is trying to execute this:
SELECT
ColumnIdentifier,
Key,
groupUniqArray(_CAST(20, 'Int32'))(Value) AS Values
FROM logs.distributed_otel_logs_kv_rollup_15m
WHERE Value != ''
GROUP BY
ColumnIdentifier,
Key
ORDER BY
ColumnIdentifier = 'NativeColumn' DESC,
ColumnIdentifier ASC,
Key ASC
LIMIT _CAST(30, 'Int32')
By removing the cast to groupUniqArray(20)(Value) AS Values this query works.
What version of ClickHouse are you on? Is there anything odd about your schema?
I've used ReplicatedSummingMergeTree to create the metadata tables and then pointed HDX to the distributed table built on top. Clickhouse version 26.3.2.3.
As reported in the log from the previous comment, when the query is ran manually it works with no problems so the problem is not with the logic of the tables.
The query version using the cast ran by the HDX code wont work on distributed tables
Activity
czephyr commentedon Apr 24, 2026
Possibly #2128 will help
knudtty commentedon May 6, 2026
That PR will get us part of the way there, I have a follow on planned that will enable those MVs to be used also for filters.
Introspection queries are really intensive, it's actually a quite difficult problem. MVs should trivialize the problem of fetching all values as part of a set. Subsets may still be difficult, but we'll introduce a toggle that allows you to still surface the "exact" values.
czephyr commentedon May 6, 2026
Great to see there is interest in making this work better. Thanks for the good work.
czephyr commentedon Jun 1, 2026
Tracking #2297 and #2376 as relevant
knudtty commentedon Jun 2, 2026
Hey @czephyr, wanted to let you know that there are some MVs you could add today to assist. It'll require you running some SQL on your CH instance and modifying your source in your team settings.
The file can be found here. #2128 is where the MVs were added, there's a video showing how you can add to the source.
czephyr commentedon Jun 3, 2026
I've added the MVs to my deployment making them
ReplicatedSummingMergeTreeand pointing the HDX metadata tables to the distributed version of them.I'm running into this error:
It looks like HDX is trying to execute this:
By removing the cast to
groupUniqArray(20)(Value) AS Valuesthis query works.czephyr commentedon Jun 15, 2026
@knudtty just tagging as a ping, this should be a super easy fix probably
knudtty commentedon Jun 15, 2026
What version of ClickHouse are you on? Is there anything odd about your schema?
czephyr commentedon Jun 15, 2026
I've used
ReplicatedSummingMergeTreeto create the metadata tables and then pointed HDX to the distributed table built on top. Clickhouse version26.3.2.3.As reported in the log from the previous comment, when the query is ran manually it works with no problems so the problem is not with the logic of the tables.
The query version using the cast ran by the HDX code wont work on distributed tables