Logging not only support journald log driver, but also support json-file log driver now, in the following documentations, there are description: "Aggregated logging is only supported using the journald driver in Docker", this is wrong. I searched "Aggregated logging is only supported using the journald driver in Docker" at https://docs.openshift.com/container-platform/3.5/welcome/index.html https://docs.openshift.org/latest/welcome/index.html There are 6 pages is wrong, maybe not limited to these pages. Document URL: https://docs.openshift.com/container-platform/3.5/install_config/aggregate_logging.html https://docs.openshift.com/container-platform/3.5/install_config/install/host_preparation.html https://docs.openshift.com/container-platform/3.5/install_config/revhistory_install_config.html https://docs.openshift.com/container-platform/3.5/welcome/revhistory_full.html https://docs.openshift.org/latest/install_config/install/host_preparation.html https://docs.openshift.org/latest/install_config/aggregate_logging.html Section Number and Name: Search "Aggregated logging is only supported using the journald driver in Docker" in the above pages. Describe the issue: "Aggregated logging is only supported using the journald driver in Docker", Logging also supports json-file log driver. Suggestions for improvement: Change "Aggregated logging is only supported using the journald driver in Docker" to "Aggregated logging uses journald as default log driver in Docker". Additional information:
@Jeff, Do you have comments about "Suggestions for improvement" part, is that OK for the improvement?
It is correct that we can support both drivers but we recommend using journald, the reasons that escape me now. We need to confirm if the statement should be something like: * We recommend journal * We support json as well. Rich can you comment?
We recommend journal because there is a lot more metadata available, but json-file is supported.
(In reply to Jeff Cantrill from comment #2) > It is correct that we can support both drivers but we recommend using > journald, the reasons that escape me now. We need to confirm if the > statement should be something like: > > * We recommend journal > * We support json as well. I think we have to support json-file until journald fully supports isolated quotas for amount of logs (curation) and rate of logging per-namespace. Once it does, we'll want to drop json-file since there are performance implications with using it (file-system buffer pollution from all the writes, number of open file handles required to scale up per-node pod counts, and probably others).