Can one of the admins verify that this patch is reasonable to test? (reply "ok to test", or if you trust the user, reply "add to whitelist")

If this message is too spammy, please complain to ixdy.

Labelling this PR as size/S

@antoineco thanks for tracking this one down. Do you think we should just use bigger default sizes? I normally run m3.medium or m3.large myself. I personally think that while it's nice to be able to try k8s out for cheap while spending a minimal amount of money, if you get a bad impression as a result that's a poor trade-off for everyone. And I think for production most people should be running the m3 class instead of the t2.

I think most AWS users are already aware of the risk involved by running steady loads on t2 class instances, and I don't believe anybody would seriously consider this class for production. However it's good to warn users because I'm certainly not the only person who'll get surprised by such a situation.

The graphs I posted show the load in a development cluster, which receives close to no traffic except the one generated by the devs. At some point we were running enough containers to make k8s components consume around 15% of the instances CPU constantly (not only the master). This worked just fine during a week and a half and suddenly broke without notice.

How to gain awareness of your cloud provider's quirks the hard way :)

LGTM.

@k8s-bot ok to test
@k8s-bot test this

pr builder appears to be missing, activating due to 'lgtm' label.

GCE e2e build/test failed for commit 926148c.

• Build Log
• Test Artifacts
• Internal Jenkins Results

@k8s-bot test this please

GCE e2e test build/test passed for commit 926148c.

• Build Log
• Test Artifacts
• Internal Jenkins Results

Automatic merge from submit-queue