Added a comment

This commit is contained in:
https://launchpad.net/~liori 2018-08-17 21:58:03 +00:00 committed by admin
parent d1eabddee6
commit 5d0c336994

View file

@ -0,0 +1,9 @@
[[!comment format=mdwn
username="https://launchpad.net/~liori"
nickname="liori"
avatar="http://cdn.libravatar.org/avatar/e1d0fdc746b3d21bb147160d40815e37b257b9119774d21784939b2d3ba95a91"
subject="comment 2"
date="2018-08-17T21:58:03Z"
content="""
So, indeed it works. However, syncing a large archive (2TB, ~1M objects) over several USB HDDs at the same time ended up a bad idea: inevitably at some point every process tried syncing with the same, slowest, repository, and lack of good I/O concurrency on a USB-connected HDD killed all performance. I'm guessing that introducing some targetted locking could probably help, or even maybe just randomizing the order in which repositories are pulled from/pushed to.
"""]]