This issue was create to adjust S2S set throttling to xmpp:15k:0:disc,bin:120m:0:disc and for other connections use xmpp:50k:0:disc,400m:0:disc@, however as I added this settings and was trying to verify this settings are applied, I found out that for cluster connections following setting is used @xmpp:25m:0:disc,bin:2000m:0:disc@. Queue overflow should not happen often in this case, so I checked @etc/init.properties file used by sure.im servers and found following entries:

--cm-ht-traffic-throttling=xmpp:60k:0:drop,bin:800m:0:drop
--cm-traffic-throttling=xmpp:100:0:disc,bin:20m:0:disc

This settings override throttling for EVERY component using HT or normal throttling. In case of first of this lines it sets xmpp:60k:0:drop,bin:800m:0:drop for S2S connections (a lot bigger than our new default value we talked about) and it also forces SAME value for CLUSTER connections!

I think this is cause of issue from #3956 - setting which limits traffic over cluster connection.

To fix this we should replace first line with:

--cm-ht-traffic-throttling=xmpp:60k:0:drop,bin:800m:0:drop
cl-comp/traffic-throttling=xmpp:25m:0:disc,bin:20000m:0:disc

or following for every component we want to change throttling limit:

comp-name/traffic-throttling=xmpp:60k:0:drop,bin:800m:0:drop

However I do not know why this setting was added? Increase limits for S2S?

If so, then reducing limit for S2S in default is not a good idea?

%eric %wojtek - do you know way this setting was introduced?

%kobit - in this case I think this issue should not be done, right?

It could have been added long time ago when we had some issues with traffic on the installation. I think you can just change it to the new defaults or comment it out, when the new code is installed.

I committed changes to repository with new defaults and changed current configuration on sure.im to use new defaults for S2S.

I left cluster without restarting, so new settings are not used yet.

To apply this new settings on sure.im, restart of cluster nodes is needed.

I noticed the cluster was restarted last night, so it should be OK now. Let's keep an eye on it. I mean our monitor keeps an eye and if that happens again we will know.