Hi. I am trying to create and delete ZFS snapshots using the crontab, creation works ok, but I am facing issues with the deleting part.

The crontab entry looks like this:

SHELL=/usr/bin/zsh

10 21 * * * /usr/sbin/zfs destroy -r tank@`TZ=$TZ+167 date +'%Y-%m-%d'` 2>> /var/log/zfs.destroy

However I am getting the following error:

zsh:1: unmatched `

If you are wondering the +167 is the maximum offset value TZ accepts, with this I attempt to delete snapshots older than 7 days, or so.

the command works just fine in root ZSH shell, but it fails in crontab, so there must be some differences, but I don't know, how to find it.

I have already tried to change the command to : tank@$(TZ=$TZ+167 date +'%Y-%m-%d') but it didn't help

Please let me know if oyu have any better solution for this.

your could move the code into a shell script file with #!/usr/bin/zsh on the first line and led crontab call that file

my apporach to deleting autmatic snapshots is having a dedidacted constant part in the name like "autobackup_" and the date in seconds. Then a zfs list -r -t snapshot -o name | grep autobackup_ and then not delete the last n-number snapshots

(or egrep "@autobackup_1[0-9]*$" )

tomww: Thanks for the hint, I'll test it and will also think about your approach.

Thanks again.

you’re welcome

I'd liek to use a USB drive for cold backup. I'd liek to have a ZFS pool on it, so i can send snapshots there. Am I supposed to create a partition and create Zpool on that partition, or should I use the raw device itself for the pool?

tomww: what do you think about this aproach:

/usr/sbin/zfs list -d 1 -H -o name -t snapshot | grep "tank@" | tail -n +7 | xargs /usr/sbin/zfs destroy -r

-d 1 means it lists non-recursively (so it emmits only the top level dataset names

-H means it ommits the output header (this is not really needed)

-o will instruct zfs to list only the name property

tail -n +7 means it ommits the first 7 items

So it deletes all snapshots except the 7 most recent ones

I try to never stript a "destroy -r". instead I try to identify every single zfs filesystem and loop over that one to find all snapshots in there.

and for the first run I prefix the destroy command with "echo" to only see the command it _would_ run.

tomw, I agree.

I have modified my script like this:

/usr/sbin/zfs list -d 1 -o name -t snapshot | grep "tank@autobackup" | tail -n +7 | xargs /usr/sbin/zfs destroy -r

note: I'm only reading occationally here, so scripting support is very limited on my side

included autobackup in the grep, so it will list only applicable snapshots from tank and will keep the 7 most recent version

I still do not use a "-r" in destroy. think of a clone dependent on a snapshot. I might be killed the clone.

I haven't tried zfs cloning feature yet.

I haven't used many of the zfs features yet.

But thanks for the hint.

if a clone is dependent on a snapshot -r is not enough to delete it. You need -R in that case.

I am wondering, why is there . and .. in my /

. is everywhere

. and .. show up in my omnios and linux systems in /