Hi, is it possible to put multiple s3 gateway ‘domain_name’ in lakefsconfig? The use case is we want to have two ways to access it: 1. via external domain name https://lakefs.xxx.xxx.xx though load balancer . 2. direct access inside k8s, usually like http://lakefs

Hi @Xubo Fei, This is unfortunately not currently possible. That said, the use-case sounds important, and I am unsure of a workaround (beyond dns hacks inside k8s, which may be possible).

Thanks Ariel

@Xubo Fei If you have some time, could you have a look at https://github.com/treeverse/lakeFS/pull/1720 ? This is what I plan to push. Reminder that host-based addressing -- which AWS S3 clients are allowed to use just about any time unless you disable it -- means that you need not only a DNS record like

<http://s3.local.lakefs.io|s3.local.lakefs.io>

but also a wildcard

*.<http://s3.local.lakefs.io|s3.local.lakefs.io>

to point at your lakefs server. (This is not due to our implementation, it is just the way of S3...).

Thanks Ariel, for us now, I need support lakefss3.aaa.bbb.com , lakefss3.aaa.local, and lakefss3

Hi! This is actually (sometimes!) not enough, due to the continuing use of host-addressing on S3 (https://aws.amazon.com/blogs/aws/amazon-s3-path-deprecation-plan-the-rest-of-the-story/). So you would also have to forward the wildcards (all suffixes)

*.<http://lakefss3.aaa.bbb.com|lakefss3.aaa.bbb.com>

,

*.lakefss3.aaa.local

,

*.lakefss3

. The last one might be trickiest, it usually uses some different DNS server. NOT all S3 applications require this, but some might. It would be an ops requirement -- not lakeFS -- but it will still be difficult.

thanks. some of the changes are in our roadmap. the last one is in k8s cluster, not through any external dns server.