search

sapia-oss / corus

Corus server
http://www.sapia-oss.org/projects/corus/index.html
GNU General Public License v3.0
2 stars 1 forks source link

Old nodes kept forever in cluster view #66

Open jcdesrochers opened 5 years ago

jcdesrochers commented 5 years ago

Old nodes are kept forever and never removed from the cluster view. As the number of node increases, the number of async thread required to perform the health check increases and flooded the log file with rejected exceptions (thread pool exhaustion).

Cluster validation should be clever and not use all the threads to perform internal validation of other peers. Pacing, throttling could be use to be scalable.

jcdesrochers commented 5 years ago

Improvement of how hosts of cluster is managed was pushed in 8ba621bbbcfaac7989c728e86000ce34db983b12

Will perform further tests before closing this issue...