With the fix for bug#865811, katello-configure now adapts the pulp.conf settings to reflect the optimal level for the system it's running on. katello-configure takes into account the number of cores available on the system.
If the pulp timeouts are addressed by automatically setting an appropriate value in pulp.conf, what is needed for documentation for customers?
If we want to document tuning the setting in pulp.conf, my impression is that might be better suited for kbase. Thoughts?
#867388
When synchronizing large channels, Pulp may return timeout errors. To keep Pulp from saturating its task queue, add the following lines to the /etc/pulp/pulp.conf file:
[tasking]
concurrency_threshold = N
N refers to double the number of CPUs less one (2 * #CPUs - 1). The recommended minimum number of CPUs is two, so the minimum recommended value for the concurrency_threshold is 3 (2x2-1). These configuration values will adjust the number of concurrent synchronizations to a level where the server is responsive to subsequent requests.