synchronous removal?

synchronous removal?

[K. Richard Pixley]

  I have an application where I want to snapshot, then do something, and 
based on the result, snapshot either the result or the previous state 
and then repeat.

So far, so good.  But eventually my disk fills and I want to remove the 
oldest snapshots, as many as I need to in order to make room enough for 
the next cycle.

I notice that when I remove old snapshots and delete old directories, 
the free space on my disk, (according to df), doesn't rise immediately.  
But instead, I see an active btrfs_cleaner for a while and my free space 
rises while it runs.  I'm presuming that the removed files and snapshots 
aren't fully reclaimed immediately but rather wait for something akin to 
a garbage collection much the way modern berkeley file systems do.

How can I either:

a) wait for the cleaner to digest the free space
b) determine that the cleaner has digested all available free space for 
now, (if not I can sleep for a while)
c) synchronously force the cleaner to digest available free space
d) something else I haven't thought of yet

Basically, I want to check to see if there's enough space available.  If 
not, I want to remove some things, (including at least one snapshot), 
wait for the cleaner to digest, and then start over with the checking to 
see if there's enough space available and loop until I've removed enough 
things that there is enough space available.  How can I do that on a 
btrfs file system?

--rich

Re: synchronous removal?

[Leonidas Spyropoulos]

I think a cron job checking the output of df could do that.
The shell script will check if there is enough space to create a snapsh=
ot
otherwise remove a snapshot.

How about that?

On Sun, Aug 1, 2010 at 10:11 PM, K. Richard Pixley <rich@noir.com> wrot=
e:
> =A0I have an application where I want to snapshot, then do something,=
 and
> based on the result, snapshot either the result or the previous state=
 and
> then repeat.
>
> So far, so good. =A0But eventually my disk fills and I want to remove=
 the
> oldest snapshots, as many as I need to in order to make room enough f=
or the
> next cycle.
>
> I notice that when I remove old snapshots and delete old directories,=
 the
> free space on my disk, (according to df), doesn't rise immediately. =A0=
But
> instead, I see an active btrfs_cleaner for a while and my free space =
rises
> while it runs. =A0I'm presuming that the removed files and snapshots =
aren't
> fully reclaimed immediately but rather wait for something akin to a g=
arbage
> collection much the way modern berkeley file systems do.
>
> How can I either:
>
> a) wait for the cleaner to digest the free space
> b) determine that the cleaner has digested all available free space f=
or now,
> (if not I can sleep for a while)
> c) synchronously force the cleaner to digest available free space
> d) something else I haven't thought of yet
>
> Basically, I want to check to see if there's enough space available. =
=A0If
> not, I want to remove some things, (including at least one snapshot),=
 wait
> for the cleaner to digest, and then start over with the checking to s=
ee if
> there's enough space available and loop until I've removed enough thi=
ngs
> that there is enough space available. =A0How can I do that on a btrfs=
 file
> system?
>
> --rich
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs=
" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at =A0http://vger.kernel.org/majordomo-info.html
>



--=20
Caution: breathing may be hazardous to your health.
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" =
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: synchronous removal?

[K. Richard Pixley]

How would you determine whether to remove another snapshot or to wait for previously removed space to be digested?

If you simply remove snapshots then you'll end up removing all if your snapshots and df will still say you don't have enough space. Btdt. What I'm doing right now is removing a snapshot and immediately sleeping for 60 seconds in hopes that it will be digested in that time. Judicious use of df and log lines tell me that some of the space is digested in that time but I have no way, (that I know of), to determine whether all of it has been.

--rich

On Aug 2, 2010, at 4:35, Leonidas Spyropoulos <artafinde@gmail.com> wrote:

> I think a cron job checking the output of df could do that.
> The shell script will check if there is enough space to create a snapshot
> otherwise remove a snapshot.
> 
> How about that?
> 
> On Sun, Aug 1, 2010 at 10:11 PM, K. Richard Pixley <rich@noir.com> wrote:
>>  I have an application where I want to snapshot, then do something, and
>> based on the result, snapshot either the result or the previous state and
>> then repeat.
>> 
>> So far, so good.  But eventually my disk fills and I want to remove the
>> oldest snapshots, as many as I need to in order to make room enough for the
>> next cycle.
>> 
>> I notice that when I remove old snapshots and delete old directories, the
>> free space on my disk, (according to df), doesn't rise immediately.  But
>> instead, I see an active btrfs_cleaner for a while and my free space rises
>> while it runs.  I'm presuming that the removed files and snapshots aren't
>> fully reclaimed immediately but rather wait for something akin to a garbage
>> collection much the way modern berkeley file systems do.
>> 
>> How can I either:
>> 
>> a) wait for the cleaner to digest the free space
>> b) determine that the cleaner has digested all available free space for now,
>> (if not I can sleep for a while)
>> c) synchronously force the cleaner to digest available free space
>> d) something else I haven't thought of yet
>> 
>> Basically, I want to check to see if there's enough space available.  If
>> not, I want to remove some things, (including at least one snapshot), wait
>> for the cleaner to digest, and then start over with the checking to see if
>> there's enough space available and loop until I've removed enough things
>> that there is enough space available.  How can I do that on a btrfs file
>> system?
>> 
>> --rich
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> 
> 
> 
> 
> -- 
> Caution: breathing may be hazardous to your health.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: synchronous removal?

[Andreas Philipp]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
 
Hi,

I came across this when I tried to build a backup server. Was there
any change on this in the meantime?

Thanks,
Andreas

On 02.08.2010 17:25, K. Richard Pixley wrote:
> How would you determine whether to remove another snapshot or to
> wait for previously removed space to be digested?
>
> If you simply remove snapshots then you'll end up removing all if
> your snapshots and df will still say you don't have enough space.
> Btdt. What I'm doing right now is removing a snapshot and
> immediately sleeping for 60 seconds in hopes that it will be
> digested in that time. Judicious use of df and log lines tell me
> that some of the space is digested in that time but I have no way,
> (that I know of), to determine whether all of it has been.
>
> --rich
>
> On Aug 2, 2010, at 4:35, Leonidas Spyropoulos <artafinde@gmail.com>
> wrote:
>
>> I think a cron job checking the output of df could do that. The
>> shell script will check if there is enough space to create a
>> snapshot otherwise remove a snapshot.
>>
>> How about that?
>>
>> On Sun, Aug 1, 2010 at 10:11 PM, K. Richard Pixley
>> <rich@noir.com> wrote:
>>> I have an application where I want to snapshot, then do
>>> something, and based on the result, snapshot either the result
>>> or the previous state and then repeat.
>>>
>>> So far, so good. But eventually my disk fills and I want to
>>> remove the oldest snapshots, as many as I need to in order to
>>> make room enough for the next cycle.
>>>
>>> I notice that when I remove old snapshots and delete old
>>> directories, the free space on my disk, (according to df),
>>> doesn't rise immediately. But instead, I see an active
>>> btrfs_cleaner for a while and my free space rises while it
>>> runs. I'm presuming that the removed files and snapshots
>>> aren't fully reclaimed immediately but rather wait for
>>> something akin to a garbage collection much the way modern
>>> berkeley file systems do.
>>>
>>> How can I either:
>>>
>>> a) wait for the cleaner to digest the free space b) determine
>>> that the cleaner has digested all available free space for
>>> now, (if not I can sleep for a while) c) synchronously force
>>> the cleaner to digest available free space d) something else I
>>> haven't thought of yet
>>>
>>> Basically, I want to check to see if there's enough space
>>> available. If not, I want to remove some things, (including at
>>> least one snapshot), wait for the cleaner to digest, and then
>>> start over with the checking to see if there's enough space
>>> available and loop until I've removed enough things that there
>>> is enough space available. How can I do that on a btrfs file
>>> system?
>>>
>>> --rich -- To unsubscribe from this list: send the line
>>> "unsubscribe linux-btrfs" in the body of a message to
>>> majordomo@vger.kernel.org More majordomo info at
>>> http://vger.kernel.org/majordomo-info.html
>>>
>>
>>
>>
>> -- Caution: breathing may be hazardous to your health. -- To
>> unsubscribe from this list: send the line "unsubscribe
>> linux-btrfs" in the body of a message to
>> majordomo@vger.kernel.org More majordomo info at
>> http://vger.kernel.org/majordomo-info.html
> -- To unsubscribe from this list: send the line "unsubscribe
> linux-btrfs" in the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
 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=jxH9
-----END PGP SIGNATURE-----

Re: synchronous removal?

[Bruce Guenter]

[-- Attachment #1: Type: text/plain, Size: 1094 bytes --]

On Sun, Aug 01, 2010 at 02:11:15PM -0700, K. Richard Pixley wrote:
> I notice that when I remove old snapshots and delete old directories, 
> the free space on my disk, (according to df), doesn't rise immediately.  

> Basically, I want to check to see if there's enough space available.  If 
> not, I want to remove some things, (including at least one snapshot), 
> wait for the cleaner to digest, and then start over with the checking to 
> see if there's enough space available and loop until I've removed enough 
> things that there is enough space available.  How can I do that on a 
> btrfs file system?

I asked a similar question a while back, and the short answer is that
you can't, short of unmounting and remounting the filesystem.  The
indication was made that writing a new ioctl to wait for all background
activity wouldn't be too hard, but I don't recall seeing it in any
recent patches.

See this thread:
http://www.mail-archive.com/linux-btrfs@vger.kernel.org/msg04872.html

-- 
Bruce Guenter <bruce@untroubled.org>                http://untroubled.org/

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: synchronous removal?

[K. Richard Pixley]

On Aug 2, 2010, at 8:02, Bruce Guenter <bruce@untroubled.org> wrote:

> On Sun, Aug 01, 2010 at 02:11:15PM -0700, K. Richard Pixley wrote:
>> I notice that when I remove old snapshots and delete old directories, 
>> the free space on my disk, (according to df), doesn't rise immediately.  
> 
>> Basically, I want to check to see if there's enough space available.  If 
>> not, I want to remove some things, (including at least one snapshot), 
>> wait for the cleaner to digest, and then start over with the checking to 
>> see if there's enough space available and loop until I've removed enough 
>> things that there is enough space available.  How can I do that on a 
>> btrfs file system?
> 
> I asked a similar question a while back, and the short answer is that
> you can't, short of unmounting and remounting the filesystem.  The
> indication was made that writing a new ioctl to wait for all background
> activity wouldn't be too hard, but I don't recall seeing it in any
> recent patches.

That's interesting. Thank you. 

Would anyone like to make a bid on doing this work?  That is, is there anyone with the knowledge and code familiarity available for me to commission this work?

--rich
>