requestTimeout in 4.x

[ich199]

Hi,

I have been experiencing an issue where Kibana (4.4.1) returns an Unknown Error when running long queries/visualisations. If the Elasticsearch cluster takes longer than 120 seconds to return the results then the error occurs. This seems to be completely unaffected by the elasticsearch.requestTimeout and elasticsearch.shardTimeout settings in the kibana.yml config file. The error appears consistently at 120 seconds.

Initial investigates seem to show that the Kibana server process is closing the connection to the web browser due to a default NodeJS timeout of 120 seconds (https://nodejs.org/docs/v0.12.10/api/http.html#http_server_timeout) as the default value for routes.timeout.socket (http://hapijs.com/api/8.8.1#route-options) is being used in the hapi server connectionOptions in https://github.com/elastic/kibana/blob/4.x/src/server/http/index.js

There also appears to be a second default hapi timeout being hit in the elasticsearch proxy (https://github.com/elastic/kibana/blob/4.x/src/plugins/elasticsearch/lib/create_proxy.js) the if this issue is fixed, of 180 seconds, as described http://hapijs.com/api/8.8.1#built-in-handlers which causes a HTTP 504 Gateway timeout error.

All the changes I have made to remove these default timeouts are in my commmt ich199@86eb9d6

[w33ble]

@ich199 thanks for the input, and for tracking down the issue.

I see you have all these changes in a local 4.x branch in your repo, but you haven't opened a pull request against that branch, or even that single commit. Would you mind opening a PR?

[ich199]

@w33ble Pull request opened as requested.

[w33ble]

Closed via #6377 and #6385

[jasonswearingen]

@ich199 thanks for posting this including the complete root cause. I don't use kibana but use hapi and was wondering why my sessions dropped at 120 seconds.