2011/10/20 Jim Klimov <jimkli...@cos.ru>:
> 2011-10-19 17:54, Fajar A. Nugraha пишет:
>> On Wed, Oct 19, 2011 at 7:52 PM, Jim Klimov<jimkli...@cos.ru>  wrote:
>>> Well, just for the sake of completeness: most of our systems are using
>>> zfs-auto-snap service, including Solaris 10 systems datiing from Sol10u6.
>>> Installation of relevant packages from SXCE (ranging snv_117-snv_130) was
>>> trivial, but some script-patching was in order. I think, replacement of the
>>> ksh interpreter to ksh93.
>> Yes, I remembered reading about that.
> Actually, I revised the systems: the scripts are kept in original form, but
> those sol10 servers where ksh93 was absent, got a symlink:
> /usr/bin/ksh93 -> ../dt/bin/dtksh

For the sake of completeness, I tried using zfs-auto-snapshot 0.12
(the last version before obsoleted by  time-slider). I used opencsw's
ksh 93 (http://www.opencsw.org/package/ksh/) and symlink
/opt/csw/bin/ksh to /usr/bin/ksh93.

It kinda works when only used for automated snapshot. But when using
zfs/backup-save-cmd property ( to make remote system do a "zfs receive
-d -F", problem started to happen. It complains the snapshot already
exist. Probably the result of interaction between
com.sun:auto-snapshot user property, child dataset, and "zfs send -R".
If only "zfs receive" has the option to "force delete the snapshot if
it already exist, rollback to the previous one, and retry the
incremental receive" :)

Since zfs-auto-snapshot has been obsoleted (and obviously
unsupported), it wouldn't make sense for me to try to fix this. I
might revisit this option again later using time-slider when solaris
11 is released though.

Thanks for the info so far.

zfs-discuss mailing list

Reply via email to