Problemsby GlyphTrouble with ClickHouse column
15 posts from 9 people11d activeFading, +0 this week
Posts per day
The posts behind it
A table created `AS` a table function whose explicit structure has other column names cannot be read
### Company or project name _No response_ ### Describe what's wrong `CREATE TABLE t (columns) AS table_function(...)` is accepted when the table function has an explicit structure whose column names differ from the columns of the table, but such a table can never be read....
Support multiple text indexes on the same column
### Use case A document-search application needs to compare two analyses of the same original text, such as literal matching and accent-insensitive matching. It should be possible to build and test a replacement index while keeping the current one available, without storing a...
MaterializedPostgreSQL: reading a Nullable column's .null subcolumn (x IS NULL, count(x)) fails with NOT_FOUND_COLUMN_IN_BLOCK on an empty table in 26.8
...and on a plain MergeTree. **Error message and/or stacktrace** **Additional context** The workaround is `optimize_functions_to_subcolumns = 0`, set in the session or settings profile of every reader of such tables. ### Version info - Resolved by: #128953 - Merged into:...
Analyzer: hashing a subquery column walks the whole subquery, GROUP BY passes over a wide CTE spend seconds in `getTreeHash`
**Describe the situation** `IQueryTreeNode::getTreeHash` of a `ColumnNode` includes the subtree of the column source (the weak pointer to the `TableNode` / `QueryNode` the column comes from). For a table that is cheap. For a subquery or CTE it is the whole subquery body. Every...
`FilterTransform` re-hashes column names on every chunk; `ExpressionTransform` is positional since #109380 but the filter is not
**Describe the situation** `FilterTransform::doTransform` builds a `Block` with `Block::cloneWithColumns` and runs `ExpressionActions::execute(Block &)`. Both index the block by name: `Block::insert` hashes every column name into `index_by_name`, and `execute` looks the inputs...
Merges load the statistics of every column from `statistics.packed` with one request per column
### Describe the situation A merge loads the column statistics of every source part before it reads any data: `MergeTask` calls `IMergeTreeDataPart::loadStatistics()` once per source part, without `required_columns`, so the statistics of every column are loaded. A mutation...
VirtualRowTransform throws NOT_IMPLEMENTED on lambda column (Code 48)
...Expected the query to return rows normally; instead it throws Code 48 NOT_IMPLEMENTED. - Workaround is setting `read_in_order_use_virtual_row = 0`. Root cause: `VirtualRowTransform.cpp:98` builds virtual-row columns via `type->createColumn()` for every header column, but...
`GROUP BY` top-K filter misses the #122486 name check: stored column reads `1`
### Describe what's wrong - `SELECT k, max(`__topKFilter(k)`) ... GROUP BY k LIMIT n` over `MergeTree` with PREWHERE returns the filter's internal value instead of the stored column. - Triggered when a real column is named `__topKFilter( )` and a PREWHERE (explicit or moved from...
Macro get_time_filter() fails in virtual datasets when the date column is not present in the final SELECT clause
...**get_time_filter()** Jinja macro to handle time filters inside the dataset, the macro fails to work correctly if the date column used in the macro is not included in the final SELECT statement. At the same time Superset query automatically appends the date column in the final...
`DROP INDEX` deletes column marks in pairs the new check admits as legal
### Describe what's wrong - A column named `skp_idx_ ` next to index ` ` loses its `.cmrk2` marks file when the index is dropped or recalculated. - Trigger: one side's file name is hashed, via a name past `max_file_name_length` or a setting change between writes. - Expected the...
Outer `USING` alias key reads left column, not inner `FULL`/`RIGHT USING` key
### Describe what's wrong - With `analyzer_compatibility_join_using_top_level_identifier = 1`, an outer `USING` key aliased over an inner `FULL`/`RIGHT JOIN ... USING` key joins wrong rows. - Triggered by `a FULL JOIN b USING (id) JOIN l USING (bucket)` where bucket aliases `id...
WITH FILL emits rows outside FROM..TO when STEP overflows the column type
### Describe what's wrong - `FillingRow::next` accepts a step that wraps the carrier type instead of rejecting it as overflow - Trigger: a fill range whose end is within one `STEP` of the type's max, e.g. `Date` near 2148/2149 - Expected 2 rows within FROM..TO but got 16509 rows...
`ALTER UPDATE/DELETE` on a column shadowing `_part`: NOT_FOUND_COLUMN_IN_BLOCK
...`validate_mutation_query = 0` makes the identical statement work, so validation itself is broken. Root cause: `getAvailableColumnsWithVirtuals` appends every virtual name to the physical names without skipping ones already present, so `_part` appears twice in `available_columns`...
Logical error: Sizes of nested column and null map of Nullable column are not equal after deserialization when reading a Wide part with a pending `MODIFY COLUMN Array(String) -> Array(Nullable(String))` of a `Nested` member (STID: 6726-6449)
🕵 Stress test (`arm_msan`) aborted the server with a new logical error, STID 6726-6449: CI report: https://s3.amazonaws.com/clickhouse-test-reports/praktika.html?PR=118757&sha=373d38db5baef55043f8767bab8f582d324f5da3&name_0=PR&name_1=Stress%20test%20%28arm_msan%29 (core dump...
Analyzer: NOT_AN_AGGREGATE for alias column usage
### Company or project name _No response_ ### Describe what's wrong NOT_AN_AGGREGATE exception when alias column is used ### Does it reproduce on the most recent release? Yes ### How to reproduce https://fiddle.clickhouse.com/12bcf9e7-e595-4e05-9c0a-cf171e050aa7 ### Expected...
Companies and products named
About this problem
Evidence
15 posts from 9 people in 2 places, about 2 a week over 44 days. Mostly on ClickHouse/ClickHouse, apache/superset. Tools named alongside: GitHub, Apache Superset, Slack, MinIO.
Frustration Frustration 0 of 3· Seen on GitHub issues
Summary, model-written
ClickHouse users encounter date range overflow bugs, duplicate virtual column errors, alias aggregate exceptions, and stress test server crashes.
How it was grouped
Posts about ClickHouse with no known issue, sharing the word "column". First post Sep 21, 2026, latest Nov 3, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method
Other ClickHouse problems
ClickHouse upgrades break working setups
31 posts from 20 people on GitHub issues.
ClickHouse users keep filing bugs
Top post: "Tighten seccomp policy for optional kernel interfaces", one of 19 posts from 8 people on GitHub issues.
ClickHouse costs outgrow the value it gives
11 posts from 3 people on GitHub issues.
Trouble with ClickHouse table
Top post: "`SYSTEM RESTORE REPLICA` on a second replica over-bumps the table `metadata` znode version; every later", one of 8 posts from 5 people on GitHub issues.
Experiments on ClickHouse are hard to reproduce and track
8 posts from 1 person on GitHub issues. Unverified.
Trouble with ClickHouse tables
Top post: "Paimon: resolve columns by field id, so tables with renamed or re-added columns are readable", one of 7 posts from 6 people on GitHub issues.
Trouble with ClickHouse join
Top post: "Explicit grace_hash now spills at max_bytes_ratio_before_external_join = 0.5 although the build side fits in", one of 5 posts from 4 people on GitHub issues.
ClickHouse error messages are cryptic
5 posts from 3 people on GitHub issues.
Trouble with ClickHouse read
Top post: "ORC: a corrupt PostScript `metadata_length` makes ORC predicate pushdown read `stripe_stats` out of bounds,", one of 4 posts from 4 people on GitHub issues.
Trouble with ClickHouse database
Top post: "`DROP NAMED COLLECTION` succeeds while a database uses the collection, and the server then fails to start", one of 4 posts from 2 people on GitHub issues.
Trouble with ClickHouse query
Top post: "Memory tracker counts jemalloc's freed pages as used, so a heavy query started right after another spills to", one of 4 posts from 3 people on GitHub issues.
ClickHouse rate limits and usage caps block work
3 posts from 3 people on GitHub issues.
History