feat: endpoint choose by health check #109
Conversation
tzssangglass
changed the title
|
@ Yiyiyimu welcome to look at this PR |
There was a problem hiding this comment.
we need to update the file https://github.com/api7/lua-resty-etcd/blob/master/rockspec/lua-resty-etcd-master-0.1-0.rockspec#L20 too
|
Add: t/v2/dir.t is unstable. |
|
@Yiyiyimu welcome to look at this PR too |
|
|
||
|
|
||
|
|
||
| === TEST 8: has no healthy etcd endpoint, follow old style |
There was a problem hiding this comment.
I didn't get why when no endpoints are healthy, we should follow old style. Since as #101 talked about
When all instances of a certain api fail (such as auth api), it will cause crazy retries, which may eventually overwhelm the ETCD cluster.
Following old styles still could not solve this problem
There was a problem hiding this comment.
I can't refute this example.
we do not request etcd when all etcd instances are down.
what do you think @membphis ?
There was a problem hiding this comment.
IMO maybe we could leave some warnings and like add some delay for continuous retries? I'm not so sure.
Hi @nic-chen do you have some opinions on this
|
ping @tzssangglass |
later, my computer is broken, will continue after repairing. |
…ty-etcd into healthcheck Conflicts: health_check.md
fix #101
fix #55