3

We had a fun issue yesterday and today where after a restart of some nodes, we ended up with more or less every pod on one server. After some research, it seems like most of the solutions for this issue are either manual or rely on something outside AKS to handle it.

Is there a way within AKS to handle the following:

  1. Balance pods across nodes regularly
  2. Ensure that multiple pods of the same service are not on the same node where possible.
4
  • Do you mean a tool directly within AKS and not using any other AWS services? Oct 16 '19 at 15:02
  • 1
    @WesleyRolnick Ideally something within Azure. The concern with adding a lot of extra 3rd party things is potential breakage when something changes, but if that's our only option, then we'll go with that. Oct 16 '19 at 15:55
  • I misread and thought this was AWS not Azure. Is a solution that adds more Azure services acceptable or are you trying to accomplish everything directly within AKS? Oct 16 '19 at 16:05
  • @WesleyRolnick Adding additional Azure services would be fine Oct 16 '19 at 18:08