Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix docker hanging when container killed #3612

Merged
merged 1 commit into from
Feb 21, 2017

Conversation

ruflin
Copy link
Member

@ruflin ruflin commented Feb 17, 2017

No timeout was passed to the docker client. It seems in case of a killed container it can happen that the connection is hanging. To interrupt this connection, the timeout from the metricset is passed to the client. That means in case info for a container cannot be fetched, it will timeout.

This change requires that the docker module is not run with a timeout of 3s seconds, which indirectly means a period of 3s. The reason is that already the http request waits ~2s for the response. So if 1s is set as timeout, all requests will timeout.

Further changes:

  • Containers without names will be ignored, as these are containers for which the data could not be fetched.
  • Period was set to 1s by default instead of the period as document. This was changed.
  • Add documentation node about minimal period.

Closes #3610

The issue with this PR was introduce in 5.2.1 by fixing the memory leak. Before go routines just piled up, but now they caused filebeat to hang.

This needs also backport to 5.2.2

@ruflin ruflin added bug Metricbeat Metricbeat needs_backport PR is waiting to be backported to other branches. review labels Feb 17, 2017
@ruflin ruflin force-pushed the fix-docker-deadlock branch 2 times, most recently from f3a071d to 80ef5c2 Compare February 21, 2017 07:10
@@ -49,6 +49,9 @@ metricbeat.modules:
cpu_ticks: true
----

It is strongly recommend to not run docker metricsets with a period smaller then 3 seconds. The request to the docker
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"It is strongly recommended"

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

fixed

Copy link
Contributor

@tsg tsg left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

No timeout was passed to the docker client. It seems in case of a killed container it can happen that the connection is hanging. To interrupt this connection, the timeout from the metricset is passed to the client. That means in case info for a container cannot be fetched, it will timeout.

This change requires that the docker module is not run with a timeout of 3s seconds, which indirectly means a period of 3s. The reason is that already the http request waits ~2s for the response. So if 1s is set as timeout, all requests will timeout.

Further changes:

* Containers without names will be ignored, as these are containers for which the data could not be fetched.
* Period was set to 1s by default instead of the period as document. This was changed.
* Add documentation node about minimal period.

Closes elastic#3610

The issue with this PR was introduce in 5.2.1 by fixing the memory leak. Before go routines just piled up, but now they caused filebeat to hang.

This needs also backport to 5.2.2
@tsg tsg merged commit 99f17d6 into elastic:master Feb 21, 2017
tsg pushed a commit to tsg/beats that referenced this pull request Feb 21, 2017
No timeout was passed to the docker client. It seems in case of a killed container it can happen that the connection is hanging. To interrupt this connection, the timeout from the metricset is passed to the client. That means in case info for a container cannot be fetched, it will timeout.

This change requires that the docker module is not run with a timeout of 3s seconds, which indirectly means a period of 3s. The reason is that already the http request waits ~2s for the response. So if 1s is set as timeout, all requests will timeout.

Further changes:

* Containers without names will be ignored, as these are containers for which the data could not be fetched.
* Period was set to 1s by default instead of the period as document. This was changed.
* Add documentation node about minimal period.

Closes elastic#3610

The issue with this PR was introduce in 5.2.1 by fixing the memory leak. Before go routines just piled up, but now they caused filebeat to hang.

This needs also backport to 5.2.2
(cherry picked from commit 99f17d6)
ruflin added a commit to ruflin/beats that referenced this pull request Feb 21, 2017
No timeout was passed to the docker client. It seems in case of a killed container it can happen that the connection is hanging. To interrupt this connection, the timeout from the metricset is passed to the client. That means in case info for a container cannot be fetched, it will timeout.

This change requires that the docker module is not run with a timeout of 3s seconds, which indirectly means a period of 3s. The reason is that already the http request waits ~2s for the response. So if 1s is set as timeout, all requests will timeout.

Further changes:

* Containers without names will be ignored, as these are containers for which the data could not be fetched.
* Period was set to 1s by default instead of the period as document. This was changed.
* Add documentation node about minimal period.

Closes elastic#3610

The issue with this PR was introduce in 5.2.1 by fixing the memory leak. Before go routines just piled up, but now they caused filebeat to hang.

This needs also backport to 5.2.2
(cherry picked from commit 99f17d6)
ruflin pushed a commit that referenced this pull request Feb 21, 2017
No timeout was passed to the docker client. It seems in case of a killed container it can happen that the connection is hanging. To interrupt this connection, the timeout from the metricset is passed to the client. That means in case info for a container cannot be fetched, it will timeout.

This change requires that the docker module is not run with a timeout of 3s seconds, which indirectly means a period of 3s. The reason is that already the http request waits ~2s for the response. So if 1s is set as timeout, all requests will timeout.

Further changes:

* Containers without names will be ignored, as these are containers for which the data could not be fetched.
* Period was set to 1s by default instead of the period as document. This was changed.
* Add documentation node about minimal period.

Closes #3610

The issue with this PR was introduce in 5.2.1 by fixing the memory leak. Before go routines just piled up, but now they caused filebeat to hang.

This needs also backport to 5.2.2
(cherry picked from commit 99f17d6)
ruflin added a commit to ruflin/beats that referenced this pull request Feb 21, 2017
No timeout was passed to the docker client. It seems in case of a killed container it can happen that the connection is hanging. To interrupt this connection, the timeout from the metricset is passed to the client. That means in case info for a container cannot be fetched, it will timeout.

This change requires that the docker module is not run with a timeout of 3s seconds, which indirectly means a period of 3s. The reason is that already the http request waits ~2s for the response. So if 1s is set as timeout, all requests will timeout.

Further changes:

* Containers without names will be ignored, as these are containers for which the data could not be fetched.
* Period was set to 1s by default instead of the period as document. This was changed.
* Add documentation node about minimal period.

Closes elastic#3610

The issue with this PR was introduce in 5.2.1 by fixing the memory leak. Before go routines just piled up, but now they caused filebeat to hang.

This needs also backport to 5.2.2
(cherry picked from commit 99f17d6)
tsg pushed a commit that referenced this pull request Feb 21, 2017
No timeout was passed to the docker client. It seems in case of a killed container it can happen that the connection is hanging. To interrupt this connection, the timeout from the metricset is passed to the client. That means in case info for a container cannot be fetched, it will timeout.

This change requires that the docker module is not run with a timeout of 3s seconds, which indirectly means a period of 3s. The reason is that already the http request waits ~2s for the response. So if 1s is set as timeout, all requests will timeout.

Further changes:

* Containers without names will be ignored, as these are containers for which the data could not be fetched.
* Period was set to 1s by default instead of the period as document. This was changed.
* Add documentation node about minimal period.

Closes #3610

The issue with this PR was introduce in 5.2.1 by fixing the memory leak. Before go routines just piled up, but now they caused filebeat to hang.

This needs also backport to 5.2.2
(cherry picked from commit 99f17d6)
@ruflin ruflin deleted the fix-docker-deadlock branch February 21, 2017 15:46
@andrewkroh andrewkroh removed the needs_backport PR is waiting to be backported to other branches. label Feb 22, 2017
leweafan pushed a commit to leweafan/beats that referenced this pull request Apr 28, 2023
No timeout was passed to the docker client. It seems in case of a killed container it can happen that the connection is hanging. To interrupt this connection, the timeout from the metricset is passed to the client. That means in case info for a container cannot be fetched, it will timeout.

This change requires that the docker module is not run with a timeout of 3s seconds, which indirectly means a period of 3s. The reason is that already the http request waits ~2s for the response. So if 1s is set as timeout, all requests will timeout.

Further changes:

* Containers without names will be ignored, as these are containers for which the data could not be fetched.
* Period was set to 1s by default instead of the period as document. This was changed.
* Add documentation node about minimal period.

Closes elastic#3610

The issue with this PR was introduce in 5.2.1 by fixing the memory leak. Before go routines just piled up, but now they caused filebeat to hang.

This needs also backport to 5.2.2
(cherry picked from commit 06ecd67)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Metricbeat stops reporting data if container is killed
3 participants