This day went from usual Thursday to worst day of my life in the span of about
10 seconds. Here''s the scenario:
2 Computer, both Solaris 10u8, one is the primary, one is the backup. Primary
system is RAIDZ2, Backup is RAIDZ with 4 drives. Every night, Primary mirrors
to Backup using the ''zfs send'' command. The Backup receives
that with ''zfs recv -vFd''. This ensures that both machines
have an identical set of filesystems/snapshots every night. (Snapshots are
taken on Primary every hour during the workday).
The issue began monday when Primary failed. After restoring it to operating
condition I began restoring the filesystems from Backup, again using ZFS
send/recv. By midnight, only about half of the data had recovered, at which
point Primary attempted its regularly schedule mirror operation with Backup.
One of our primary ZFS filesystems had not yet been restored, and since it
wasn''t on Primary when the mirror operation began, ''zfs
recv'' destroyed it on the Backup system. AHHHHH.
So, in short, a RAIDZ array contained 7 ZFS filesystems + dozens of snapshots in
one RAIDZ pool. 12 hours ago some of those filesystems were destroyed,
effectively by a zfs destroy command (executed by zfs recv). No data has been
written to that pool since then. Is there anyway to revert it to the state it
was in 12 hours ago?
--
This message posted from opensolaris.org
