Package: qa.debian.org Severity: wishlist User: qa.debian....@packages.debian.org Usertags: udd
Dear UDD heroes, the publicly available udd.dump [1] is 1.5 GB and I'd like to ask for the possibility to turn it from a PGDMP in `pg_dump -Fc` format to a dir'ed `pg_dump -Fd` which might, with the many tables within udd, provide a measure of bandwidth-savings for mirrors of the database when the periodic dumps don't update some of the tables and can therefore be skipped in a mirror run. (Potentially could save time on dumps and restores as well, as -Fd can dump and restore tables in parallel.) Further saving (and ease of mirror use) could possibly be achieved by providing an rsync module of the dump directory [1], if the possibility exists of setting that up on ullmann.d.o. Reasoning: Since the database is constantly changing and may require a mirror to update often, grabbing a new 1.5GB file perhaps multiple times daily can be daunting to some lower-bandwidth connections (while some tables actually get updates only daily or less). After a chat with mapreri I don't have high hopes wrt the rsync module, but I was encouraged to just put it out there anyway. < mapreri> nyov: of course, if you wish a -Fd dump, it should be provided alongside the current -Fc one, not instead. And that is probably the final nail in the coffin - having two having both is really a necessity. As an alternative option -- perhaps udd.dump could be dumped without pg_dump internal compression (-Z0) and instead piped through lzma/xz [2], as the popcon dump already does? [1] https://udd.debian.org/dumps/ [2] https://salsa.debian.org/qa/udd/blob/master/scripts/dump-db-more-frequently https://salsa.debian.org/qa/udd/blob/master/scripts/dump-db.sh Cheers
