Use multiple Tigris instances to run tests #1878
Conversation
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## main #1878 +/- ##
==========================================
- Coverage 66.11% 65.95% -0.17%
==========================================
Files 312 313 +1
Lines 15230 15240 +10
==========================================
- Hits 10070 10051 -19
- Misses 4163 4194 +31
+ Partials 997 995 -2
Flags with carried forward coverage won't be shown. Click here to find out more. |
| // getNextTigrisPort gets the next port number of Tigris to be used | ||
| // for testing in Round Robin fashion. |
integration/setup/startup.go
Outdated
| ) | ||
|
|
||
| // ports are available port of Tigris. | ||
| var ports = []uint16{8081, 8082, 8083, 8085, 8086} |
There was a problem hiding this comment.
I like this idea, but I'm a bit skeptical about these ports being defined here as the exact number of exact ports and not being configurable.
Is there a plan to make those things configurable (like tigris-url itself) - maybe in separate PRs?
There was a problem hiding this comment.
I've looked at it briefly and decided I would not do it for this, currently there were several places where ports are defined, env tool, docker-compose.yml, and within integrations tests. So it's nice to have configurable but also it touches more than the integration test parameter. For this PR, I'm gonna do quick win. 🙈
There was a problem hiding this comment.
Oh, envtool is a nice catch as we wait for 8081 only, and not for the other ports...
I'm ok to merge it as is to have a quick win, and maybe we can reconsider this and make it configurable after #1857 (so, it'll be easier to deal with configuration).
There was a problem hiding this comment.
I think we should start paying a bit more attention to our code comments and not leave explanations on GitHub only, where they are unlikely to be found.
In this case, cross-linking those ports to the Docker Compose configuration would significantly improve understanding.
There was a problem hiding this comment.
Yeah agree. I've added comments. I also foresee the number of instances are just a guess and need adjusting hence a new issue #1887
cmd/envtool/envtool.go
Outdated
| if err != nil { | ||
| return err | ||
| } | ||
| // Skip 8084 because since it is used by userland proxy. |
There was a problem hiding this comment.
It seems to fail only on CI. https://github.com/FerretDB/FerretDB/actions/runs/4040272813/jobs/6945764486
I updated comment to something a bit more understandable.
Error response from daemon: driver failed programming external connectivity on endpoint ferretdb_speedup_3 (331153047b6e2c6218d96d6ac924a2831bdda68f069b8f8a364d00670cc1484f): Error starting userland proxy: listen tcp4 0.0.0.0:8084: bind: address already in use
task: Failed to run task "env-up-detach": exit status 1
chilagrow
dismissed stale reviews from w84thesun, rumyantseva, and noisersup
via
8f906a4
Use different ports instead of documenting problems. Do not copy&paste the same comment; use references instead. Do not link to closed issue.
AlekSi
dismissed stale reviews from rumyantseva, noisersup, and w84thesun
via
bd05d83
AlekSi
changed the title
|
Thank you for the clean up @AlekSi 🤗 |
Description
Closes #1568.
Use 4 additional instances for running tests for tigris exposed on ports 8082, 8083, 8085 and 8086. Skipping 8084 because it's used by something else.
On CI, this change appears to runs tigris tests in 6 to 7 minutes. I checked recent merges to main which tigris tests appear to run in 10 to 11min. I don't know how much more $$ it might use.
Readiness checklist
task all, and it passed.@FerretDB/core), Assignee, Labels, Project and project's Sprint fields.