-
Notifications
You must be signed in to change notification settings - Fork 871
Open
Labels
1 Bug1 Feature3 AQLQuery language relatedQuery language related3 Index3 OptimizerAQL query optimizer is involvedAQL query optimizer is involvedWaiting User Reply
Description
My Environment
- ArangoDB Version: 3.7.11
- Storage Engine: RocksDB
- Deployment Mode: Cluster
- Deployment Strategy: Kubernetes
- Configuration:
networkAttachedVolumes: trueon theArangoDeployment,--server.statistics=trueas extra arg - Infrastructure: own
- Operating System: Ubuntu 20.04.2
- Total RAM in your machine: 3 nodes, all 64GB
- Disks in use: 3 nodes, all SSD 512GB
- Used Package: Docker - official Docker library (
arangodb/arangodb:3.7.11)
Component, Query & Data
Affected feature:
UPSERT index usage
Size of your Dataset on disk: 20GB
Replication Factor & Number of Shards (Cluster only):
all collections have replicationFactor: 3, writeConcern: 2 Number of shards: 1
Steps to reproduce
Problem:
I have 2 identical edge collections (same setting, same indexes). The same UPSERT query:
UPSERT {
_from: 'a',
_to: 'b'
}
INSERT {}
UPDATE {}
IN collection
RETURN {}
uses the correct (best) index for the first collection but uses the worse one for the second collection (see image above)
I thought maybe I could provide an indexHint but that doesn't appear to be possible in UPSERT and probably should be.
Expected result:
either
- both pick the best index automatically, or
UPSERTallows to provide anindexHint
Metadata
Metadata
Assignees
Labels
1 Bug1 Feature3 AQLQuery language relatedQuery language related3 Index3 OptimizerAQL query optimizer is involvedAQL query optimizer is involvedWaiting User Reply
