[dev mode] Kibana regularly considers (wrongly) that all source files have changed and continuously restart

[fbaligand]

Kibana version:
master
Elasticsearch version:
master
Server OS version:
Windows 10

Original install method (e.g. download page, yum, from source, etc.):
dev mode using yarn start --oss
Describe the bug:
When I start Kibana in dev mode, regularly (maybe every 30 minutes), Kibana considers wrongly that all source files have changed and continuously restart.
Once it arrives, I have to stop Kibana, delete "src" dir, then execute git reset --hard, and finally start Kibana.
This is quite annoying and painful to develop on Kibana.
I would say I get this behaviour since Kibana 6.2.

Expected behavior:
Detect only files that are really modified.

Provide logs and/or server output (if relevant):

 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\get_panel_data.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\get_table_data.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\table\build_request_body.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\build_processor_function.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\request_processors\table\index.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\request_processors\table\pivot.js"
 restarting server  due to changes in "src\core_plugins\metrics\common\basic_aggs.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\helpers\get_buckets_path.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\helpers\bucket_transform.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\helpers\parse_settings.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\helpers\parse_interval.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\request_processors\table\query.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\helpers\get_timerange.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\get_interval_and_timefield.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\request_processors\table\split_by_everything.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\request_processors\table\split_by_terms.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\request_processors\table\date_histogram.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\helpers\get_bucket_size.js"
 restarting server  due to changes in "src\core_plugins\metrics\server\lib\vis_data\helpers\calculate_auto.js"

[ReneU]

I have the same issue with 3 different windows machines (using git bash). None of the files is really modified. Thanks for the workaround.

[tsullivan]

ping @elastic/kibana-operations

[ReneU]

I figured out that all my machines having this issue only have 4GB RAM. Everything works fine with 8GB or more.

[fbaligand]

My machine has 8 GB RAM, and I still have the issue.
The only way I found to avoid the issue is to set « —no-watch » CLI flag when I start kibana

[spalger]

Is there a chance you're using Dropbox or something that might be interacting with your Kibana repo other than git/IDE? What IDE are you using?

We might be able to enhance the logging to describe why the change was detected, and ignore if it we can't figure out the reason.

[fbaligand]

Hi @spalger
My folder is "C:\kibana" and is not a Dropbox share (or any other cloud shared folder).
My IDE is Eclipse.
Concerning IDE, I had already Eclipse IDE when I got Kibana 6.1, and didn't have this behavior.

Thanks for your try to understand the issue cause!

[spalger]

We don't have access to the reason why the files are considered changed unfortunately, but it seems this could quite possibly be caused by this: nodejs/node#21643

nodejs/node#21643 (comment) recommends disabling the "last accessed timestamp" to fix the issue, and the original poster suggests that it works.

Would either of you like to give this a shot?

[jbudz]

Closing this out as stale - lets reopen if there's further issues.