Re: [mlmmj] mlmmj patches from distributions: Gentoo
Baptiste Daroussin <[email protected]> Fri, 09 Dec 2022 05:25:58 +0000
| Newsgroups | org.mlmmj.mlmmj |
|---|---|
| Message-ID | <[email protected]> |
Le 9 décembre 2022 00:18:46 GMT+01:00, "Robin H. Johnson" <[email protected]> a écrit : >On Thu, Dec 08, 2022 at 09:01:29AM +0100, Baptiste Daroussin wrote: >> > There's one scaling discussion we need to have, but the handling needs >> > to include how to incrementally get there: >> > >> > How to have millions of mails in /archive/! >> > >> > Our most active list is now approaching 1.5M emails in that directory, >> > and it worries me. >> >> Why is it worrying you? a directory can hold way more than that. >> >> On FreeBSD, one thing we are doing is when creating the public archives via: >> https://fossil.nours.eu/mlmmj-archiver/doc/trunk/README.md (this may move to >> codeberg), I append the archives to a mbox, so I can cleanup what ever is in the >> archives directory if I need (I did not need up to now). >Cleaning up the archives directory will break the +get-NNNN >functionality. > >It makes the directory take extremely long to scan during backups. >I think it should be partitioned more, but any +get-NNNN code will need >to support both partitioned & non-partitioned. > Good point and easy to implement, I ll looking into it. >mbox support for +get would be nice, but that's much more work than just >partitioning the directory. > >We do have our own archives site: >https://archives.gentoo.org/ > >The older versions used to use mhonarc, but also ran into scaling >concerns, so we built something more custom (earlier versions used >ElasticSearch as a DB; I don't recall what the current version uses >underneath, but we can re-ingest everything if we need to). > If this is public I am interrested into it, i wrote mlmmj-archiver for scaling issues as well. Best regards, Bapt