journal log file

journal log file

[JPASTORM]

Hi,

I'm running reiserfs with a linux 2.2.20 kernel. Where does the system
create the journal log file for each reiserfs partition? I have currently 3
partitions (/, /var and /cache), and two of them were created with reiserfs
(/var and /cache). The thing is my / partition hasn't a lot of room to keep
the journal log file. What's the biggest size it can reach? Will I have one
log file for each reiserfs partition? Thanks in advance.

Juan Pastor
E-mail: jpastorm@esade.edu

Centro de Proceso de Datos
Fundación ESADE
Av. Pedralbes, 60-62
08034 Barcelona

Tel  +34 932806162 ext. 2562
Fax  +34 932048105

Re: journal log file

[Oleg Drokin]

Hello!

On Sat, May 18, 2002 at 12:30:26PM +0200, JPASTORM@esade.edu wrote:

> I'm running reiserfs with a linux 2.2.20 kernel. Where does the system
> create the journal log file for each reiserfs partition? I have currently 3

Reiserfs does not have journal "file".
In 2.2 kernel series (reiserfs v3.5) journal is placed at the beginning
of the partition explicitly (and occupies 32M).

> partitions (/, /var and /cache), and two of them were created with reiserfs
> (/var and /cache). The thing is my / partition hasn't a lot of room to keep
> the journal log file. What's the biggest size it can reach? Will I have one

For a filesystem with 4k blocksize standard journal is always 32M (and variable
blocksize or relocated journal patches won't be produced for 2.2 kernels
anyway).

> log file for each reiserfs partition? Thanks in advance.

Yes, each reiserfs filesystem have its own journal.

Bye,
    Oleg

Re: journal log file

[Dieter Nützel]

On Monday 20 May 2002 06:15, Oleg Drokin wrote:
> > partitions (/, /var and /cache), and two of them were created with
> > reiserfs (/var and /cache). The thing is my / partition hasn't a lot of
> > room to keep the journal log file. What's the biggest size it can reach?
> > Will I have one
>
> For a filesystem with 4k blocksize standard journal is always 32M (and
> variable blocksize or relocated journal patches won't be produced for 2.2
> kernels anyway).

Oleg, you should have mentioned that Chris had a "variable" log size patch for 
"ages" but he only haven't had enough time for polishing it.

I for myself were very excited when we get this in some day 'cause than I have 
some little more space on my "boot" partition left or I can shrink it 
finally.

Regards,
	Dieter

Re: journal log file

[Chris Mason]

On Mon, 2002-05-20 at 10:02, Dieter Nützel wrote:
> On Monday 20 May 2002 06:15, Oleg Drokin wrote:
> > > partitions (/, /var and /cache), and two of them were created with
> > > reiserfs (/var and /cache). The thing is my / partition hasn't a lot of
> > > room to keep the journal log file. What's the biggest size it can reach?
> > > Will I have one
> >
> > For a filesystem with 4k blocksize standard journal is always 32M (and
> > variable blocksize or relocated journal patches won't be produced for 2.2
> > kernels anyway).
> 
> Oleg, you should have mentioned that Chris had a "variable" log size patch for 
> "ages" but he only haven't had enough time for polishing it.
> 
> I for myself were very excited when we get this in some day 'cause than I have 
> some little more space on my "boot" partition left or I can shrink it 
> finally.

Note, namesys guys wrote this patch, it is in 2.5.x already.

-chris

Re: journal log file

[Oleg Drokin]

Hello!

On Mon, May 20, 2002 at 04:02:21PM +0200, Dieter N?tzel wrote:
> > For a filesystem with 4k blocksize standard journal is always 32M (and
> > variable blocksize or relocated journal patches won't be produced for 2.2
> > kernels anyway).
> Oleg, you should have mentioned that Chris had a "variable" log size patch for 
> "ages" but he only haven't had enough time for polishing it.

Not for 2.2 kernel.
And the question was about 2.2

> I for myself were very excited when we get this in some day 'cause than I have 
> some little more space on my "boot" partition left or I can shrink it 
> finally.

You seems to be brave enough to use 2.5 kernels, it seems ;)

Bye,
    Oleg

Re: journal log file

[Dieter Nützel]

On Monday 20 May 2002 16:45, Oleg Drokin wrote:
> Hello!
>
> On Mon, May 20, 2002 at 04:02:21PM +0200, Dieter N?tzel wrote:
> > > For a filesystem with 4k blocksize standard journal is always 32M (and
> > > variable blocksize or relocated journal patches won't be produced for
> > > 2.2 kernels anyway).
> >
> > Oleg, you should have mentioned that Chris had a "variable" log size
> > patch for "ages" but he only haven't had enough time for polishing it.
>
> Not for 2.2 kernel.
> And the question was about 2.2

OK, you win...;-)

> > I for myself were very excited when we get this in some day 'cause than I
> > have some little more space on my "boot" partition left or I can shrink
> > it finally.
>
> You seems to be brave enough to use 2.5 kernels, it seems ;)

Maybe :-)))

Cheers,
	Dieter

Re: journal log file

[Oleg Drokin]

Hello!

On Sat, May 25, 2002 at 12:59:40PM +0200, JPASTORM@esade.edu wrote:

> After the mkreiserfs and mount commands so as to convert ext2 partitions
> into reisrerfs, I noticed some diferences between the du and df commands
> output.
> [root@apollo /var]# df
> Filesystem           1k-blocks      Used Available Use% Mounted on
> /dev/sda5              7168784   1637808   5530976  23% /cache
> /dev/sda6               461788    145536    316252  32% /var
> [root@apollo /cache]# du
> 1591088 .

33*1024(journal size + some reserved space)+1591088 = 1624880
and 16k were probably occupied by internal tree nodes, that are not
accounted for files, so du do not show this.

> [root@apollo /var]# du
> 114596  .

33*1024+114596 = 148388
Here we see that actual number of bytes is less than we expected.
This is because of tail packing and du's feature.
du only takes into account amount of block each file occupies.
So if two files occupy the same block, du would count amount of blocks
used to be 2, where correct amount is 1 (tail packing).

Probably you have some amount of small files on your /var partition.
(smal files are files that are less than 16K in size).

> The df value for /var partition was more or less expected having in count
> these 32 MB of the journal. Could you explain me the difference in the
> /cache partition (it's just the squid cache partition)?

I do not see much difference, in fact.

> I also understand that I cannot make a reiserfs partition from an ext2
> partition if I have less than these 32 MB left in it and I want to preserve
> all data.

There is no transparent tool to convert ext2 fs to reiserfs.
The current procedure is to backup all the data, reformat the partition,
and then to restore the data back.
And yes, the 33M would be claimed for journal and some other internal to
filesystem stuff.
Also it is highly advisable to have at least 10% of free space on the partition,
so that large-scale fragmentation would less likely to occur.

Bye,
    Oleg