Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 8 Nov 2007 22:53:32 +0300
From:      "Artem Kuchin" <matrix@itlegion.ru>
To:        "Yuri Pankov" <yuri.pankov@gmail.com>
Cc:        freebsd-current@freebsd.org, Pawel Jakub Dawidek <pjd@freebsd.org>
Subject:   Re: gjournal weirdness contd
Message-ID:  <002001c82241$0b722df0$0c00a8c0@Artem>
References:  <046501c82190$815ecca0$0c00a8c0@Artem><001701c8223d$df3aeef0$0c00a8c0@Artem> <20071108194655.GD18314@darklight.org.ru>

next in thread | previous in thread | raw e-mail | index | archive | help
Yuri Pankov wrote:
> On Thu, Nov 08, 2007 at 10:30:47PM +0300, Artem Kuchin wrote:
>> All of the below is still there. I hope someone would help
>> with this. Pawel, please, answer.
>> 
>>> FReebSD 7-BETA2, amd64 (as of nov 5th)
>>> Here is what i did this time:
>>> WEIRDNESS N1:
>>> fdisk ad6
>>> lable ad6 - only one fs for the full disc
>>> now i have /dev/ads1d
>>> then i do
>>> gjournal load
>>> gjournal label -f  /dev/ad6s1d
>>> and now i get
>>> /dev/ad6
>>> /dev/ad6s1
>>> /dev/ad6s1.journal
>>> /dev/ad6s1.journalc
>>> /dev/ad6s1.journald
>>> What are all journal and why c and d are put AFTER
>>> the 'journal' word?
>>> omni3# gjournal list
>>> Geom name: gjournal 1827255885
>>> ID: 1827255885
>>> Providers:
>>> 1. Name: da0s1f.journal
>>>   Mediasize: 304473987584 (284G)
>>>   Sectorsize: 512
>>>   Mode: r0w0e0
>>> Consumers:
>>> 1. Name: da0s1f
>>>   Mediasize: 306473988096 (285G)
>>>   Sectorsize: 512
>>>   Mode: r1w1e1
>>>   Jend: 306473987584
>>>   Jstart: 304473987584
>>>   Role: Data,Journal
>>> Geom name: gjournal 3462346836
>>> ID: 3462346836
>>> Providers:
>>> 1. Name: ad6s1.journal
>>>   Mediasize: 397938328064 (371G)
>>>   Sectorsize: 512
>>>   Mode: r0w0e0
>>> Consumers:
>>> 1. Name: ad6s1
>>>   Mediasize: 400085812224 (373G)
>>>   Sectorsize: 512
>>>   Mode: r1w1e1
>>>   Jend: 400085811712
>>>   Jstart: 397938328064
>>>   Role: Data,Journal
>>> da0 is a twa backed gjournal fs. And it is fine.
>>> ad6  is just connected to motherboard sata controller.
>>> journal size is approx 2GB (see weirdness 2)
>>> As you see, it seemslike gjournla is attached direcly to
>>> ad6s1 (TO SLICE!!!). I don't get it, why so many journal?
>>> why partition letter is at the end? What to mount?
>>> WEIDNESS N2:
>>> If i want to specify 2GB journal size
>>> i put
>>> label -s  2147483648
>>> and after than i get 'journal size is too small'
>>> if i do
>>> label -s  2048000000
>>> it works
>>> So, i patched the source a bit and set 2GB by default :)
>>> BUt this is all weird.
> 
> I've seen the same behaviour with gjournal. Reason, as I understand
> it, is sysinstall creating first partition on slice at offset 0 and
> not 63 
> as gjournal expects. I've manually edited disklabel and changed offset
> of first partition to be 63 (and, of course, decreasing partition
> size). And I guess you see complaints from gjournal about this on
> console :-) 
> 
> So now I have:
>  a: 67108801       63    4.2BSD        0     0     0
> 
> and no more weird behaviour from gjournal.
> 
> BTW, it was really funny to observe it. Stopping gjournal on ad4s1a
> with `gjournal stop` would lead to gjournal now thinking it is on
> ad4s1 or ad4s1c.

Aha! Then i'll try just  creating a very small partition at the beginning and
the the main partition. That should fix this and wasting 10Megs is not
a problem for a 400GB HDD :)

Yuri, have you ever tried to specify exctly 2GB journal size? like
gjournal label -s  2147483648

For me it says "journal size is too small".

--
Artem





Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?002001c82241$0b722df0$0c00a8c0>