I was wondering why IOs went with their own retry configuration object instead of making FluentBackoff[1] public. Some examples are SnsIO[2] and SolrIO[3]. Was it because we thought that IOs would likely need specialized retry configuration that a general retry configuration class wouldn't apply?
1: https://github.com/apache/beam/blob/c3bd4854e879da65060de8cd259865a9b34742c7/sdks/java/core/src/main/java/org/apache/beam/sdk/util/FluentBackoff.java#L30 2: https://github.com/apache/beam/blob/da9e17288e8473925674a4691d9e86252e67d7d7/sdks/java/io/amazon-web-services2/src/main/java/org/apache/beam/sdk/io/aws2/sns/SnsIO.java#L262 3: https://github.com/apache/beam/blob/da9e17288e8473925674a4691d9e86252e67d7d7/sdks/java/io/solr/src/main/java/org/apache/beam/sdk/io/solr/SolrIO.java#L225 On Wed, Apr 15, 2020 at 11:59 AM Akshay Iyangar <[email protected]> wrote: > Hi > > > > I actually wanted a way to configure FluentBackoff at the client side for > S3 in that effort I created below PR. > > But as luke mentioned in the PR FluentBackoff is part of util and I can > directly expose it to public. > > > > So a suggested alternative was to use a Configuration class that is public > facing which then convert’s it to the internal beam class and have it > generic enough to be used across IO’s. > > > > Just wanted to know what the community feels and if the above suggestion > by luke is ok with other’s I’ll try to implement that instead. > > > > > > JIRA - 9742 > > https://github.com/apache/beam/pull/11396 > > > > Thanks > > Akshay I >
