Stacking LDS updates cause non-deterministic Envoy memory usage

[phylake]

We regularly see 3+ total_listeners_draining in Envoy stats. Each of these has a full set of filter chain matches and consume quite a bit of memory for our 200+ FQDNs

Regardless of any changes in Envoy (see envoyproxy/envoy#7923), Contour should ensure LDS doesn't stack up in Envoy. There should be a guarantee of, for example, 2 ingress_http (at most 1 draining) + 2 ingress_https (at most 1 draining) listeners as an upper bound

[davecheney]

Thank you for raising this issue. I think the best solution to this is FDS which we’re waiting on. Maybe it’ll come in envoy 1.12. With that said I understand the issues that frequent LDS update cause in your environment and I will continue to work to reduce them. I don’t have a concrete eta on when this will happen or what it will involve but 499 and related issues remain a high priority for me.

…

On 5 Oct 2019, at 02:13, Brandon Cook ***@***.***> wrote:  We regularly see 3+ total_listeners_draining in Envoy stats. Each of these has a full set of filter chain matches and consume quite a bit of memory for our 200+ FQDNs Regardless of any changes in Envoy (see envoyproxy/envoy#7923), Contour should Ensure LDS doesn't stack up in Envoy. There should be a guarantee of, for example, 2 ingress_http (at most 1 draining) + 2 ingress_https (at most 1 draining) listeners as an upper bound — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[github-actions]

The Contour project currently lacks enough contributors to adequately respond to all Issues.

This bot triages Issues according to the following rules:

• After 60d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, the Issue is closed

You can:

• Mark this Issue as fresh by commenting
• Close this Issue
• Offer to help out with triage

Please send feedback to the #contour channel in the Kubernetes Slack

[github-actions]

The Contour project currently lacks enough contributors to adequately respond to all Issues.

This bot triages Issues according to the following rules:

• After 60d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, the Issue is closed

You can:

• Mark this Issue as fresh by commenting
• Close this Issue
• Offer to help out with triage

Please send feedback to the #contour channel in the Kubernetes Slack