So our DBs are all like 150 - 200gb, majority is InnoDB. Backup and restore process takes a good 4 - 6 hours.
I was making some new slaves and used InnoBackupEx like usual. I ended up wasting a couple days as sometimes my .frm/.myi/.myd etc files didn’t get included in the backup.
So I’d start up my DB and wouldn’t see any error messages, but doing a mysqlrepair would only repair the mysql. db and not any of the other ones. Using mysql client and doing a ‘use ; show tables;’ I would get an empty set.
I was using Xtrabackup 1.2 on some hosts but have upgraded to xtrabackup 1.4 for Debian and have been running it like this:
innobackupex -user=xxxx -passwordxxxx --defaults-file=/blah/blah/my.cnf --slave-info /mnt/pathtonfs/
then from same host do
innobackupex -user=xxxx -passwordxxxx --defaults-file=/blah/blah/my.cnf --apply-log /mnt/pathtonfs/2010-12-12_20-23-12/
I then place the into the data directory
The only difference I can think of is that some of my MySQL (vanilla) servers I tried running the backup on are masters with high load and some of them are running 5.1.38 (haven’t wanted to bring them down to upgrade). The rest are using 5.1.51 or 5.1.52 – binary distros.
I’m migrating these servers to Percona 5.1.52 running Ubuntu.
I know this post doesn’t really give that much information – I’m just wondering if there’s something I could be doing wrong? And is there anyway to get the .frm/.myd/.myi after the fact? Would simply rsyncing these over work?