use realistic coturn memory limits#50
Conversation
mastaab
force-pushed
the
coturn-memory-limits
branch
from
b399248 to
849b993
Compare
|
| memory: 256Mi | ||
| limits: | ||
| memory: 6.5Gi | ||
| memory: 512Mi |
There was a problem hiding this comment.
I'm not sure we should be as agressive. A single coturn pod usually owns its node anyway and we should leave it up to the operator to set mem limits to whatever the node can tolerate.
There was a problem hiding this comment.
The main reason for this change is, that this limit is setting false expectations and lead to false sizing assumptions.
In prod, there is no coturn pod that is using more than 150mb. If a customer is sizing their coturn nodes, and they are driven by the limits here, the will be way too big and are just inflating the compute requirements (which especially on-prem need to be really tailored to what we need)
One can always overwrite the limits of course, but from my perspective the standard should reflect the reality.
Wdyt?
2 participants
previous memory limits where necessary because of memory leaks.
coturns should never use this amount of memory, setting saner limits now