cockroachdb / cockroachdb/cockroach
distsql: heterogeneous big endian - little endian clusters can produce incorrect results
Open
A-s390x
A-sql-execution
C-cleanup
docs-known-limitation
T-sql-queries
- Dominant language
- Go
- Stars
- 32.5k
- Forks
- 4.1k
- PR merge metrics
- PR metrics pending
Description
DistSQL machinery relies on having the same `hash` computation across all nodes in the distributed plan. As it turns out (see #146108), there is a difference at least for the hash utilities used by the vectorized engine between big endian and little endian systems. As a result, a heterogenous cluster that has nodes running on both big endian and little endian systems can produce incorrect results for distributed plans. A possible workaround if someone really needs this setup would be to disable distsql via `SET CLUSTER SETTING sql.defaults.distsql = off;`.
Jira issue: CRDB-51797
Contributor guide
Assessment
This issue has not been assessed yet.