Add support for knet transport

Bug #2055428 reported by Felipe Alencastro
14
This bug affects 3 people
Affects Status Importance Assigned to Milestone
OpenStack HA Cluster Charm
Triaged
Wishlist
Unassigned

Bug Description

Since the release of corosync 3.0 the default transport method is knet(2) and udp/udpu are still there only for backwards compatibility(3), and their maintenance is restricted to bugfixes(1).

Among the benefits of knet over udp I would highlight these three:

1: Better tolerance for small network outages(3).
 - We've had some issues multiple times, and we had to lock our machines on MAAS in order to avoid hacluster+masakari-monitors unnecessarily rebooting compute-nodes, that's an extra risk when you have nova-compute and ceph-osd on the same host.
2: Better performance and lower latencies(3).
3: Allow for multiple rings, up to 8(2-3), enables lp:1850822.
 - This would allow to have ring0 on oam-space and ring1 on internal-space, which could help with those small network hiccups

1:https://<email address hidden>/msg12806.html
2:https://people.redhat.com/ccaulfie/docs/KnetCorosync.pdf
3:https://build.clusterlabs.org/corosync/presentations/2017-Kronosnet-The-new-face-of-corosync-communications.pdf

Changed in charm-hacluster:
importance: Undecided → Wishlist
status: New → Triaged
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.