This update addresses the following:
1) Fix an issue with peer grpc connection after a vcore instance
is recycled.
2) Set the default peer grpc retry to 0 for improved API performance in a
cluster
3) Adjust the logs level in the global_handler and local_handler.
This basically removes unnecessary and performance impacting
info logs and set some logs to warning instead of info.
4) Add a test to detect whether the cluster is in a transient state.
During that time, the coordinator leader waits until all previously
running voltha instances are registered in Consul before upading the
core assignments. This means that to a previously running vcore will
remain in the assignment list during the transient period, hence some
API requests targetted that vcore will fail during that time. Once
all expected voltha instances are registered the coordinator leader
will update the core assignment key, hence subsequent API requests
should pass.
Change-Id: I918b6af8f9158444d9e612c490972493b9609344
diff --git a/voltha/voltha.yml b/voltha/voltha.yml
index 8b5b1ee..88c0b9f 100644
--- a/voltha/voltha.yml
+++ b/voltha/voltha.yml
@@ -63,9 +63,9 @@
membership_watch_relatch_delay: 0.1
membership_maintenance_loop_delay: 5
tracking_loop_delay: 1
- session_time_to_live: 300
- session_renewal_loop_delay: 30
- session_renewal_timeout: 60
+ session_time_to_live: 60
+ session_renewal_loop_delay: 10
+ session_renewal_timeout: 10
worker:
time_to_let_leader_update: 5