-
Feature Request
-
Resolution: Not a Bug
-
Critical
-
None
-
devex docs #201 May 6-May 27
-
2
-
Undefined
-
---
-
---
What is the problem that your customer is facing?
Currently there is no way to set the keepalive in the fluent forward config when forwarding logs to external fluentd.
By default keepalive is set to false, which means a new socket is created for every chunk sent. I feel that this is inefficient and would prefer to use the existing socket connection to send logs.
Especially since with the keepalive set to false, fluentd doesn't seem to close the open sockets, even if they're not being used anymore. They only get closed when the eventual network tcp timeout is reached.
What are your expectations for this feature
Enable keepalive by default.
Have you done this before and/or outside of support and if yes, how? (Optional)
We have tested this using the legacy method of using the config map to configure forwarding and it works the way we expect it to.
- documents
-
-
- Closed
-
[RHDEVDOCS-2750] Document the add keepalive option for fluent forward config
rdlugyhe screenshot looks right, but as rhn-support-hgomes says we don't want to encourage use of the legacy method.
For the CLF, there's nothing to document, it is take care of automatically when CLF generates fluentd config.
rhn-engineering-aconway Please have a look at the attached screenshot, above, and let us know if that should work.
rhn-support-hgomes Oh wow! Now I see what you did. Does that configuration file (screenshot) you just shared work?
Sri, rhn-support-hgomes: I just talked with Alan - He says the keepalive option does not appear in the ClusterLogForwarder CR YAML. It's a background setting and not something the user configures.
rhn-support-srr please close this Jira 2750, and work with Alan to add the release notes text to https://issues.redhat.com/browse/LOG-1186 and, from there, to the 5.1 release notes.
i mean, if the intention is not to talk about the keepalive setting (because it will become true as default) . Maybe we should just update the examples? and perhaps mark a numbered circle on it that explains the purpose?
rhn-engineering-aconway I see that you want to state the problem as "currently the logging collector is creating more socket connections than necessary".
- Where exactly do you want to see this update?
PTAL: The link to the related OCP doc page is here: https://docs.openshift.com/container-platform/4.7/logging/cluster-logging-external.html
For doc purposes I would state the problem as "currently the logging collector is creating more socket connections than necessary" and not mention the keepalive setting. This is the symptom that the users observes, and we want to get away from talking about fluentd internals as much as possible.
No doc task other than Release Notes, which will be handled in
LOG-1186. RN text: "currently the logging collector is creating more socket connections than necessary".Hence closing this ticket.