jslocinski

Hey, sorry, we've been so busy on the region upgrades and UI mockups that forgot to update here :) We plan to release: Gateway (DVR), Floating IPs and LoadBalancer (Octavia)...

Hello, yes, we plan (as region upgrade is progressing) to release the first shot in: GRA9 | SBG7 | UK1 | DE1 | WAW1 | BHS5 | SYD1 and probably...

I'm glad to announce that since last week, we have new services in General Availability for managing your network and traffic in Public Cloud: Load Balancer, Gateway, and Floating IPs!...

Hi @cambierr, at this moment we support local (in terms of region) Public Cloud VMs in setups with Gateway (so including pub2priv loadbalancing). We're working to extend the scope (incl....

Hi pqarmitage, when we were testing it with lower number of instances, it was always last 9 that went properly to FAULT, once all the others were not touched.

I confirm, with --enable-netlink-enobufs it works for 15 instances. But when I launch 100 instances then I see only last 18-20 is handled properly. It reminds me some old issue...

I just compiled old stable 1.2.24 (which we were playing before with many instances) to verify how does it work on the same OS/kernel (ubuntu/4.9.*) as we play recently with...

This is interesting. Hmm.. is there a simple way we can increase the buffers or split sockets to test 100 instances v4 + 100 v6? Don't think will need more...

Thanks, will try the patch soon. BTW. because of many instances and even bigger test setups we already have core/rmem_max value at the level of 64MB...

WORKS for 100 instances and 10MB buffer in keepalived. 64MB max in OS: /proc/sys/net/core/rmem_default and rmem_max: 67108864 10MB for keepalived: head /run/conf/keepalived.conf vrrp_netlink_cmd_rcv_bufs 10485760 vrrp_netlink_monitor_rcv_bufs 10485760 Mar 14 16:36:36 bz-perflab-4-vr...