Details
-
Bug
-
Resolution: Fixed
-
Major
-
6.5.0
-
Untriaged
-
Unknown
Description
While adding a test case to reproduce failures observed in Mad-Hatter system testing due to the bounce of the ns_server web server, I'm using the following to restart the web server on n_0 in a cluster_run environment:
curl -v http://couchbase:couchbase@localhost:9000/diag/eval -d "spawn(fun menelaus_sup:restart_web_servers/0)"
|
Intermittently the start of the webserver fails, with:
[menelaus:critical,2019-10-09T13:22:45.555-07:00,n_0@172.22.0.2:<0.16248.6>:menelaus_web:http_server:99]Failed to start web service: {error,eaddrinuse}
|
Granted the webserver is not intended to be bounced like this, but assuming this API can ever be called by production code in a legitimate fashion, it seems bad that we shut down the http server but are unable to restart it; any subsequent attempts to reach the web server on this node fails until couchbase server is completely bounced.
Attached are logs for n_0 which contains an instance of the failed bounce whose failure message was pasted above.