qdrant-tenant-scaling
This skill explains how to architect Qdrant for multiple tenants at scale. It covers payload-based filtering for smaller deployments, custom sharding to localize tenant data across 100k+ tenants, and tiered multitenancy for uneven tenant sizes. It also clarifies when strict isolation via separate collections is justified and what practices to avoid.
qdrant-tenant-scaling guides you through multi-tenant scaling strategies, from payload filtering for ~10k tenants to custom sharding for 100k+.
AI-generated summary based on this skill's SKILL.md
Install
qdrant/skills/tenant-scaling · repository language: Python
git clone https://github.com/qdrant/skills
cp -r skills/skills/qdrant-scaling/scaling-data-volume/tenant-scaling ~/.claude/skills/tenant-scalingnpx skillfed install qdrant/skills/tenant-scalingFrequently asked questions
AI-generated answers based on this skill's SKILL.md and metadata
How to scale Qdrant with multiple tenants without one collection per tenant?
qdrant-tenant-scaling addresses this by recommending payload-based filtering for smaller deployments, where all tenants share a single collection and queries filter by tenant_id in the payload. For larger deployments, custom sharding localizes tenant data across cluster nodes, avoiding the overhead of managing thousands of separate collections. This approach scales to 100k+ tenants while maintaining performance.
What are qdrant-tenant-scaling's tenant isolation strategies?
qdrant-tenant-scaling covers payload partitioning for basic isolation, custom sharding for data localization across nodes, and tiered multitenancy to handle uneven tenant sizes efficiently. For compliance-heavy use cases, it also explains when strict isolation via dedicated collections or encrypted shards is justified. The skill clarifies trade-offs between isolation strength and operational complexity.
How does qdrant-tenant-scaling optimize performance for uneven tenant sizes?
qdrant-tenant-scaling introduces tiered multitenancy, which assigns large tenants to dedicated shards or collections while smaller tenants share pooled resources. This prevents large tenants from starving smaller ones and reduces query latency variance. The approach balances isolation benefits with resource efficiency for deployments with highly skewed tenant workloads.
Can qdrant-tenant-scaling implement custom sharding to localize tenant data?
Yes. qdrant-tenant-scaling explains custom sharding strategies that distribute tenant data across cluster nodes based on tenant_id hashing or explicit placement rules. This localizes data, reduces network hops, and enables tenant-specific replication policies. The skill covers practical implementation patterns for deployments scaling to 100k+ tenants across distributed infrastructure.
When should qdrant-tenant-scaling recommend strict tenant isolation?
qdrant-tenant-scaling advises strict isolation—via separate collections, dedicated shards, or encryption—for compliance-driven scenarios (HIPAA, GDPR) and when tenants require encryption keys or access controls that payload filtering cannot enforce. For most other cases, it recommends payload-based or custom-sharded approaches to reduce operational overhead while maintaining adequate isolation.
What multitenancy practices does qdrant-tenant-scaling advise against?
qdrant-tenant-scaling cautions against creating one collection per tenant for 100k+ tenant deployments, as this creates unsustainable operational and memory overhead. It also warns against relying solely on payload filtering for strict compliance isolation and against ignoring uneven tenant size distribution, which can degrade performance for smaller tenants sharing resources with large ones.
SKILL.md
rendered from the published skill — quoted content, verbatim
What to Do When Scaling Multi-Tenant Qdrant
Do not create one collection per tenant. Does not scale past a few hundred and wastes resources. One company hit the 1000 collection limit after a year of collection-per-repo and had to migrate to payload partitioning. Use a shared collection with a tenant key.
- Understand multitenancy patterns Multitenancy
Here is a short summary of the patterns:
Number of Tenants is around 10k
Use the default multitenancy strategy via payload filtering.
Read about Partition by payload and Calibrate performance for best practices on indexing and query performance.
Number
(truncated - see the full file via the links below)
Read as markdown · JSON record · Browse the source repository
File tree — 1 file
skills/qdrant-scaling/scaling-data-volume/tenant-scaling/SKILL.md