I’m working on replacing an aging mysql slave server with some nice new hardware. The last step is ensuring that the backups run (and run well) on the new box, and i’m getting a little stuck there.
Old Server -
Ubuntu Server 14.04.5
5.6.35-81.0-log Percona Server (GPL), Release 81.0, Revision c96c427
xtrabackup version 2.2.12 based on MySQL server 5.6.24 Linux (x86_64) (revision id: 8726828)
New Server -
Ubuntu Server 16.04.3
5.6.37-82.2-log Percona Server (GPL), Release 82.2, Revision d1eb51005df
xtrabackup version 2.3.9 based on MySQL server 5.6.24 Linux (x86_64) (revision id: fde0e3e)
We have a ~1.3TB dataset. I was generally seeing 12-16 hour incremental backup runs (apply-log included) on the old server, and on the new server I’m seeing much longer run times. Both are/were writing to the same NFS-mounted endpoint.
I don’t have logs from the old runs, but I can and will provide backup/prepare logs from the most recent runs.
I’m also using --innodb-track-changed-pages, but that doesn’t seem to be helping at all. Also am passing --use-memory=48GB, but that doesn’t seem to change anything compared to not passing that param at all.