[Fleet] Add retry when retrieving agent version #186459
Conversation
nchaulet
added
bug
|
Pinging @elastic/fleet (Team:Fleet) |
🤖 GitHub commentsExpand to view the GitHub comments
Just comment with:
|
|
|
||
| export async function getLatestVersion(): Promise<string> { | ||
| const response: any = await axios('https://artifacts-api.elastic.co/v1/versions'); | ||
| return last(response.data.versions as string[]) || '8.1.0-SNAPSHOT'; | ||
| return pRetry(() => axios('https://artifacts-api.elastic.co/v1/versions'), { |
There was a problem hiding this comment.
The artifacts-api is not stable, don't use it. We had to move away from it in the agent team.
The latest snapshot for each version is available from https://snapshots.elastic.co which does not use the artifacts API.
Examples:
- https://snapshots.elastic.co/latest/8.15.0-SNAPSHOT.json
- https://snapshots.elastic.co/latest/8.14.2-SNAPSHOT.json
The version to use will be the current version of Kibana in a branch for the next release.
Those will give you a build ID you can use to build the artifact download URL For example 8.15.0-SNAPSHOT gives you "build_id" : "8.15.0-72200925" as the latest build ID which gets used as shown below to get the linux x64_64 .tar.gz artifact directly.
There was a problem hiding this comment.
@cmacknz Looks like a bunch of Kibana tests use the artifacts API, do you think even with retry? (and a fallback value) it's not good enough?
I am worried when using https://snapshots.elastic.co/ we will have issues when creating new branch for new version and when the elastic-agent SNAPSHOT is not yet available
There was a problem hiding this comment.
I am worried when using https://snapshots.elastic.co/ we will have issues when creating new branch for new version and when the elastic-agent SNAPSHOT is not yet available
Then we might have to pin the exact version indeed. Retrying might help but stability is not guaranteed.
There was a problem hiding this comment.
Then we might have to pin the exact version indeed. Retrying might help but stability is not guaranteed.
The way I implemented this is with retry and a fallback value to 8.15.0-SNAPSHOT if retries failed
There was a problem hiding this comment.
You can certainly start with a retry and see how it works, if it continues to be flaky the solution is as described above - don't use the artifacts-api.
💛 Build succeeded, but was flaky
Failed CI StepsTest Failures
Metrics [docs]
To update your PR or re-run it, just comment with: cc @nchaulet |
kibanamachine
added
v8.15.0
backport:skip
Summary
Resolve #184681
The test to run standalone agents are sometime failing when retrieving the agent version for the artifacts API.
That PR should fix it, by adding retries and a fallback value.