[
https://issues.apache.org/jira/browse/ARROW-16521?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17534579#comment-17534579
]
Carl Boettiger commented on ARROW-16521:
----------------------------------------
I'm seeing timeouts on s3$ls() for example, though this also impacts most other
operations. (I'm testing against a Redhat CEPH system which speaks the S3 API,
not against master Amazon itself, which no doubt explains the greater latencies
here... this probably is never an issue if I stuck with the commercial giant
but just trying to test the open source options....)
I've found that setting the env var,
{code:java}
Sys.setenv(CURLOPT_ACCEPTTIMEOUT_MS=50000) {code}
seems to work though (I think), so maybe we'll all good...
> [C++][R] Configure curl timeout policy for S3
> ---------------------------------------------
>
> Key: ARROW-16521
> URL: https://issues.apache.org/jira/browse/ARROW-16521
> Project: Apache Arrow
> Issue Type: Bug
> Components: R
> Affects Versions: 7.0.0
> Reporter: Carl Boettiger
> Priority: Major
>
> Is it possible for the user to increase the timeout allowed on the curl
> settings when accessing S3 records? The default setting appears to be more
> aggressive than most other S3 clients I use, which means that I see a lot
> more failures on arrow-based operations than the other clients see. I'm not
> seeing how this can be increased though?
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
