I'm tired of having to explain the concepts of our load balancing to our customers. Let's write a chapter in our Administrators Guide that describes our load balancing on a high level. Explain rating, penalty points, and how you can modify bogomips_per_user and ram_per_user to modify load balancer behavior. Explain when load balancer will probably do the wrong thing, and how to adjust it to work properly (clusters with hardware of differing capacity).
> MUST > > * How load balancer works must be described in the documentation. A new section has been added under the "Clustering" chapter (which also was promoted to a top-level chapter). > SHOULD > > * Considerations and tips regarding load balancing should be included in the documentation. References were added to other sections about limiting users per agent, and putting agents into "draining" mode. A note about unevenly matched agents were added (note that this will likely be fixed by bug 8566).
Went through what's been added and removed, and generally it reads perfectly fine, with just a few minor potential adjustments. The new section covers key aspects of the simplified load balancer and links to closely related information. Acceptance criteria: > MUST > * How load balancer works must be described in the documentation. ✅︎ It is described. > SHOULD > * Considerations and tips regarding load balancing should be included in the documentation. ✅︎ Check. References to relevant further reading are included. Reopening for minor adjustments.
Issues from comment 8 have been addressed. Marking as resolved.
The minor issues mentioned above have been addressed, and acceptance criteria are fulfilled! Closing.
We should probably get rid of the configuration variable "load_update_cycle" now that we don't monitor load in the same way.
I meant to reopen bug 4429
