Closed geriux closed 6 months ago
From your comment, I tried that in https://github.com/wordpress-mobile/gutenberg-mobile/pull/6699 and it doesn't work for our issues. When looking for solutions, the most mentioned approaches were to increase fetch-retry-mintimeout
and fetch-retry-maxtimeout
. For example in this issue from the npm
client repo. I also saw other posts like this one.
It appears that fetch-timeout
specifies the maximum time npm will wait for a single network request to complete before timing out. Meanwhile, the fetch-retry-mintimeout
option sets the minimum timeout value for retry attempts when npm encounters a failed network request, as seems to be the case in our CI environment due to limited network speed/stability.
I'm going to retry all jobs again just to make sure they pass the npm
installation step. 🤞
Some of the jobs will fail due to the existing artifacts in CI. I think we could try this approach and see how it performs. What do you think @fluiddot ?
Some of the jobs will fail due to the existing artifacts in CI. I think we could try this approach and see how it performs. What do you think @fluiddot ?
Sounds good, the changes won't break the CI jobs so let's try it.
Ok it is still failing 😅 . We should probably fix the CI issues by adding cache to see if it solves this npm failures.
I'll close this for now.
@geriux Another option we could try is to reduce the number of parallel connections with maxsockets
. Maybe npm
is initiating too many requests and the Buildkite instance has some kind of limitation.
@geriux Another option we could try is to reduce the number of parallel connections with
maxsockets
. Maybenpm
is initiating too many requests and the Buildkite instance has some kind of limitation.
Interesting, let's try that!
Seems changing maxsockets didn't address the issue 😞. At this point, I think we should investigate the Buildkite instance to avoid connection timeouts.
Seems changing maxsockets didn't address the issue 😞. At this point, I think we should investigate the Buildkite instance to avoid connection timeouts.
Yup, I'm trying one more thing just out of curiosity which is to remove the --prefer-offline
flag, since there's no cache at the moment. But yeah we should investigate what's happening with Buildkite instead.
This PR attempts to solve the current errors when installing
npm
dependencies in CI, mostly happening in the Android jobs with the error:It increases the following config:
fetch-retry-mintimeout
to120000
fetch-retry-maxtimeout
to300000
For comparison in my local setup, the default values are
10000
forfetch-retry-mintimeout
and60000
forfetch-retry-maxtimeout
To test all CI checks should pass.
PR submission checklist: