issues: 1433576351
This data as json
| id | node_id | number | title | user | state | locked | assignee | milestone | comments | created_at | updated_at | closed_at | author_association | pull_request | body | repo | type | active_lock_reason | performed_via_github_app | reactions | draft | state_reason |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1433576351 | I_kwDOBm6k_c5VcqOf | 1880 | Datasette with many and large databases > Memory use | 525934 | open | 0 | 4 | 2022-11-02T18:10:27Z | 2022-11-16T17:50:29Z | NONE | > Datasette maintains an in-memory SQLite database with details of the the databases, tables and columns for all of the attached databases. The above is from the docs ^. There's two problems here - the number of datasette "instances" in a single server/VM and the size of the database itself. We want the **opposite** of in-memory, including what happens on SQLlite - documented in https://www.sqlite.org/inmemorydb.html From the context in https://github.com/simonw/datasette/issues/1150 - does it mean datasette is memory-bound to the size of the dataset - which might be a deal-breaker for many large-scale use cases? In an extreme case - let's say a single server had 100 SQLlite databases, which would enable 100 "instances" of datasette to run, one per client (e.g. in a SaaS multi-tenant environment). How could we achieve all these goals: 1. Allow any _one_ of these 100 databases to grow to say 2Tb in size 2. Have one datasette instance, which connects to 1 of the 100 instances, based on incoming credentials/tenant ID 3. Minimize memory use entirely - both by datasette and SQLlite, such that almost all operations are executed in real-time on-disk with little to no memory consumption per-tenant, or per-database. Any ideas appreciated - we're looking to use this in a SaaS type of setting - many instances, single server. @simonw great work on datasette, in general! Possibly related to https://github.com/simonw/datasette/issues/1480 but we don't want use any kind of serverless infra - this is a long-running VM/server. | 107914493 | issue | {"url": "https://api.github.com/repos/simonw/datasette/issues/1880/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0} |