rero / rero-mef / 34224228377

87%

push

github

perf(docker): speed up the dev and test stack

Ports rero-ils 8ef67d4. The shard count it lowers with an index
template lived in the mapping files here, where no template can reach
it: settings stated in a create request beat every template.

* move the shard and replica count out of the twelve mapping files
  into one `record` template, and take the replicas from 2 to 1
* bring the dev and test indices down to one shard and no replica with
  an order 100 template: the `init` service installs it for the stack,
  tests/conftest.py puts it back after wiping the templates
* fix the flaky tests. The suite asked a 512 MB heap to carry 288
  shards, Elasticsearch died while creating the indices and every test
  in the module errored on a reset connection. Now 12 shards: six
  consecutive clean runs, ~125 s down to ~103 s
* relax every PostgreSQL durability guarantee, here and not in
  docker-services.yml, which is the base of the production example
* add healthchecks to the four services, and a `gate` depending on
  them and on `init`, so plain `docker compose up -d` blocks until the
  stack is usable and returns 1 rather than hanging. Replaces
  docker/wait-for-services.sh, unreferenced since .travis.yml went
* keep 8 shards for production, more than the data asks for: 45.7M
  documents measure ~18.5 GB, so the largest index holds 0.80 GB per
  shard. Lowering it needs a reindex

Co-Authored-by: Peter Weber <peter.weber@rero.ch>
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

2 of 2 new or added lines in 1 file covered. (100.0%)

5466 of 6261 relevant lines covered (87.3%)

0.87 hits per line