sporadic git command failed "git write-tree" about serverless-plugin-git-variables HOT 15 CLOSED

@petergaultney interesting, this isn't something I've seen myself - I'll take a look

from serverless-plugin-git-variables.

Can confirm that this happens to me too – although on every run, not just on the first one.

Unlike @petergaultney, I’m using Windows.

from serverless-plugin-git-variables.

Actually, I think this is related to this snippet:

I tried rewriting the snippet locally to use sequential await instead of map & Promise.all, and the error stopped happening.

Let me submit a PR.

from serverless-plugin-git-variables.

Seeing this intermittently (but frequently) on MacOSX Mojave, too.

from serverless-plugin-git-variables.

When it will be fixed???

from serverless-plugin-git-variables.

Release as 3.3.2

from serverless-plugin-git-variables.

Maybe it is not actually resolved? I already upgraded to 3.3.2 and still see this:

[2019-07-14T23:37:47.166Z]   Error --------------------------------------------------

[2019-07-14T23:37:47.166Z]  

[2019-07-14T23:37:47.166Z]   Command failed: git write-tree

[2019-07-14T23:37:47.166Z] fatal: Unable to create '/home/jenkins/workspace/s_XXXXXXX-services_develop/.git/index.lock': File exists.

[2019-07-14T23:37:47.166Z] 

[2019-07-14T23:37:47.166Z] Another git process seems to be running in this repository, e.g.

[2019-07-14T23:37:47.166Z] an editor opened by 'git commit'. Please make sure all processes

[2019-07-14T23:37:47.166Z] are terminated then try again. If it still fails, a git process

[2019-07-14T23:37:47.166Z] may have crashed in this repository earlier:

[2019-07-14T23:37:47.166Z] remove the file manually to continue.

BTW, I run several sls deploy concurrently from the same git repo in different directories for different Lambda functions.

A backoff-then-retry may resolve the issue, I think.

from serverless-plugin-git-variables.

BTW, I run several sls deploy concurrently from the same git repo in different directories for different Lambda functions.

Yeah, if you're running these in parallel outside the plugin it would still be possible to have the failure.

A backoff-then-retry may resolve the issue, I think.

It would be very nice if the plugin would handle this. More to the point, it seems weird to me that this is not a read-only operation.

What if instead of doing git write-tree and then git diff-index, the plugin did either _exec('git status --porcelain 2>/dev/null| egrep "^(M| M)" | wc -l') > 0 (for both uncommitted and untracked) or simply _exec('git diff --stat') !== '' (for uncommitted but ignoring untracked files)? Then this would neatly skip the entire issue even in parallel deployment situations.

https://stackoverflow.com/questions/2657935/checking-for-a-dirty-index-or-untracked-files-with-git

from serverless-plugin-git-variables.

@petergaultney I like this suggestion - I'd like to avoid your first idea because that makes a number of assumptions about the environment (and would fail on Windows in particular), but the second feels very reasonable. I'll have to think a bit about whether I want to release this as a breaking change or not but I'm earmarking a bit of time next week to plow through some of the issues that have piled up in this project

from serverless-plugin-git-variables.

I've tried out the PR referenced above from @alanning and it doesn't seem to solve the core problem, which is that the underlying implementation to get the git dirty status is not read-only with respect to git itself.

I'm considering trying to fix this in our own fork. If/when I do, I'll make a PR.

from serverless-plugin-git-variables.

Hi @jacob-meacham we are still facing this issue. For more info, pls see PR: #55

from serverless-plugin-git-variables.

@jacob-meacham could we re-open this? The issue is still unsolved and there are two PRs related to this one.

from serverless-plugin-git-variables.

@james-hu sure thing - I've scheduled some time this week to take a look. Thanks!

from serverless-plugin-git-variables.

@jacob-meacham I've created a new pull request for this #62. Could you have a look and see if this can be merged?

from serverless-plugin-git-variables.

Released a fix in 4.1.0. Thanks!

from serverless-plugin-git-variables.