Major Description
Based on TS-2145, I'm going to introduce cluster-wide logging stats, so that we can see the logging status of whole cluster.
I have submitted a patch, which has been tested in our product and master branch. Here is an example to show how this patch works:
I tested it with:
1) Two log clients, test78/test79, in the same cluster.
2) One log server, test81.
I used jtest to do stress testing(rm -rf /var/run/trafficserver/*.snap to clear history value before testing), after a while, stop jtest, and use following commands to verify the logging metrics:
==Log client1: test78==
[root@test78 ~]# lynx -dump http://localhost:8080/stat/ | grep -E 'cluster.*sent_to_network|log_access_ok'
proxy.node.log.event_log_access_ok=6898925
proxy.cluster.log.event_log_access_ok=8795630
proxy.cluster.log.num_sent_to_network=8795630
proxy.cluster.log.num_lost_before_sent_to_network=0
proxy.cluster.log.bytes_sent_to_network=1268834448
proxy.cluster.log.bytes_lost_before_sent_to_network=0
proxy.cluster.log.bytes_sent_to_network_avg_10s=0
proxy.process.log.event_log_access_ok=6898925
==Log client2: test79==
[root@test79 ~]# lynx -dump http://localhost:8080/stat/ | grep -E 'cluster.*sent_to_network|log_access_ok'
proxy.node.log.event_log_access_ok=1896705
proxy.cluster.log.event_log_access_ok=8795630
proxy.cluster.log.num_sent_to_network=8795630
proxy.cluster.log.num_lost_before_sent_to_network=0
proxy.cluster.log.bytes_sent_to_network=1268834448
proxy.cluster.log.bytes_lost_before_sent_to_network=0
proxy.cluster.log.bytes_sent_to_network_avg_10s=0
proxy.process.log.event_log_access_ok=1896705
==Log server: test81==
[root@test81.cn8 trafficserver]# lynx -dump http://localhost:8080/stat/ | grep -E "received_from_network|_to_disk" | grep process
proxy.process.log.num_received_from_network=8795630
proxy.process.log.num_flush_to_disk=8795630
proxy.process.log.num_lost_before_flush_to_disk=0
proxy.process.log.bytes_received_from_network=1268834448
proxy.process.log.bytes_flush_to_disk=857743696
proxy.process.log.bytes_lost_before_flush_to_disk=0
proxy.process.log.bytes_written_to_disk=857743696
proxy.process.log.bytes_lost_before_written_to_disk=0
According these stats data above, We can easy to verify whether the logs in the whole cluster are lost.