Full peers list growing excessively large (> 100K)

[stellardynamic]

Expected behavior

Node should store a reasonably sized List of peers.

Actual behavior

Nodes on several versions are storing over 100k peer entiries, in their sets, many of which are even doubles due to disconnecting and reconnecting activity of attackers.

Steps to reproduce

It appears these sets are also growing rapidly still. Here are the results of our open API nodes that have become sick after attaining a peers set this large. The issue can be fixed by rebuilding the node in most cases but still this should not be happening. I've also included other (unknown owner) open API nodes' results on testnet with those IPs half censored to demonstrate which versions this appears to be affecting.

https://test-01.liskapi.io/api/peers?limit=100&offset=100

{"meta":{"offset":100,"limit":100,"count":111040}, "version":"2.1.3-rc.0" (12 hours ago)

{"meta":{"offset":100,"limit":100,"count":140157},"version":"2.1.3-rc.0" (Most recent call)

https://test-02.liskapi.io/api/peers?limit=100&offset=100

{"meta":{"offset":100,"limit":100,"count":141534},"data":[{"version":"2.1.3-rc.0", }
(This one also expanded approx 20k or more over night.

Here are some other pubic API nodes results

{"meta":{"offset":0,"limit":1,"count":104920},"data":[{" version":"2.0.1-rc.0"

{"meta":{"offset":0,"limit":1,"count":121263},"data":[{"version":"2.1.3-rc.0"

{"meta":{"offset":0,"limit":1,"count":127036},"data":[{""version":"2.1.1-rc.0"

{"meta":{"offset":0,"limit":1,"count":120085},"data":[{"version":"2.1.3-rc.0"

{"meta":{"offset":0,"limit":1,"count":141497},"data":[{"version":"2.1.3-rc.0"

{"meta":{"offset":0,"limit":1,"count":125670},"data":[{"version":"2.1.3-rc.0"

Which version(s) does this affect? (Environment, OS, etc...)

For certain 2.0.1, 2.1.2, and 2.1.3. I do also confirm finding a 2.1.2 version nodewith the problem.

Special Thanks to Przemer and LiskPoland.PL for playing a major part in this report.