mirror of https://github.com/OpenVidu/openvidu.git
0b210fb24f
The failure clock kept running during a master outage, so when a master answered again the failure already looked older than the limit and the node was replaced at once, before LiveKit could reconnect to the restored master. Seen live on AWS elastic (ov-cloud-tester master-guard): after a 20 min master outage both Media Nodes were replaced within 17 s of the master coming back. Now no failure is counted while no master answers a Redis PING, so LiveKit gets the whole window once a master is back. |
||
|---|---|---|
| .. | ||
| output.tf | ||
| tf-gpc-openvidu-ha.tf | ||
| variables.tf | ||
| versions.tf | ||