Occasionally I see request timeout at phase/query+fetcher/scroll
. I think it's because the cluster is busy and the request didn't come back in 5 minutes (the default scroll timeout) and I see in log that the request was under path: /_search/scroll, params: {scroll=5m}
.
To verify this I want to increase the request timeout, but I don't find a parameter in the spark support page to specify scroll timeout. There's a es.http.timeout
but that value is default to 1m.
Is that the parameter I should use or is there another one?
Thank you very much!