Delay with logrotate causes degradation
Closed this issue · 0 comments
The current image of docker-metacat does a call to logrotate in docker-entrypoint.sh which logrotates catalina.out. If catalina.out is large enough (current max is 100M) then it is truncated and compressed into a gz file. This can cause a delay in start up.
During the 11/4/2020 deployment of this image to production, degradation seemed to occur with private packages as they no longer seemed to appear on ESS-DIVE. An example of this is @FiannaOBrien's package: https://data.ess-dive.lbl.gov/view/doi:10.15485/1668055, which had bad permissions issues when she tried to access it.
There were also issues found with data-dev where files to a dataset could not be uploaded which resulted in an error:
Session is required to WRITE to Node
A quick fix is to restart the metacat container. Since catalina.out already got logrotated, this process is quick and the container moves forward as usual.
Our current solution involves having logrotate run as a background process.