quickwit

quickwit copied to clipboard

- Investigate why decommissioning takes way more than `commit_timeout_secs + Δ` - Investigate why decommissioning led to many shards being open in the cluster - Indexer should process pending merges

guilload

This PR uses content-length to make sure we account the memory of the REST body upfront. It also "shrinks to fit" the buffer that is passed around.

fulmicoton

`Index XXX not found` even though the index has been created.

5

To reproduce, `run_tests.py` with commenting out `time.sleep(...)`. The ready API returns true before Quickwit is actually ready.

fulmicoton

``` ---- rate_limiter::tests::test_rate_limiter_acquire stdout ---- thread 'rate_limiter::tests::test_rate_limiter_acquire' panicked at quickwit-common/src/rate_limiter.rs:173:9: assertion failed: !rate_limiter.acquire_bytes(ByteSize::kb(20)) stack backtrace: 0: _rust_begin_unwind 1: core::panicking::panic_fmt 2: core::panicking::panic 3: quickwit_common::rate_limiter::tests::test_rate_limiter_acquire 4: quickwit_common::rate_limiter::tests::test_rate_limiter_acquire::{{closure}} 5: core::ops::function::FnOnce::call_once note: Some details...

fulmicoton

We should add the `version-revision` as an info metrics.

fulmicoton

Reorganize directories

1

We could organize by role - ingester - indexer - searcher Durability nature - temporary data - caches - storages

fulmicoton

Observed on cicada : On control plane restart, upon reception of the chitchat state, we end up removing a lot of shards. ``` 2024-04-18T12:42:53Z app[e82d41ec03e498] iad [info]2024-04-18T12:42:53.820Z INFO quickwit_control_plane::model: removing...

fulmicoton

local file ingestion success, 0 published documents/splits

5

**Describe the bug** I am trying to ingest about 1.2B documents (1.8GB x 1900 files) using single node configuration deployed on Amazon ECS. `quickwit tool local-ingest` seems successful, but no...

rnawhale

add a test to `rest-api-tests` and document

PSeitz