[9fans] fossil/flfmt -v problem

[9fans] fossil/flfmt -v problem

[andrey mirtchovski]

[ very long story short, i'm trying to recover from couple days ago's venti,
using a score i got doing 'main: vac /archive/2003/0701' ]

here's the problem:


# fossil/flfmt -v e24ddc5eaea97b3ad947c9d5a3b5a796c548249b /dev/sdD0/fossil
fs header block already exists; are you sure? [y/n]: y
48cbf63405e14adef602c44fd7e14cd6b5471793
0c273af08c6045a9ddc58840622453929e06c206
e457e4fdd6fb49bf7d4b932568aa964adc139ce1
fatal error: bad root: no qidSpace
#

Re: [9fans] fossil/flfmt -v problem

[jmk]

On Thu Jul  3 18:51:22 EDT 2003, mirtchov@cpsc.ucalgary.ca wrote:
> [ very long story short, i'm trying to recover from couple days ago's venti,
> using a score i got doing 'main: vac /archive/2003/0701' ]
>
> here's the problem:
>
>
> # fossil/flfmt -v e24ddc5eaea97b3ad947c9d5a3b5a796c548249b /dev/sdD0/fossil
> fs header block already exists; are you sure? [y/n]: y
> 48cbf63405e14adef602c44fd7e14cd6b5471793
> 0c273af08c6045a9ddc58840622453929e06c206
> e457e4fdd6fb49bf7d4b932568aa964adc139ce1
> fatal error: bad root: no qidSpace
> #
>

Flfmt needs a score created by an archive snapshot, not a vac file
(as I understand it).

It's funny, I was working on making this work (i.e. using a vac score
to initialise a filesystem) earlier in the week and 'sort of' have it
working, but there are details needing to be done that I won't get to
for a while. The particular detail you are sticking over here is that
the max qid used in the vac file is held in a different place in a 'real'
vac file (qid entry of the root directory entry) than in an archive
snapshot (root metadata qidspace entry).

Re: [9fans] fossil/flfmt -v problem

[andrey mirtchovski]

On Thu, 3 Jul 2003 jmk@plan9.bell-labs.com wrote:

> Flfmt needs a score created by an archive snapshot, not a vac file
> (as I understand it).
>

I read this in the man page, that's why I decided to go sufficiently
back in time that I wouldn't hit anything in the fossil snaps (which
go back 23 hours).

My understanding was that the 'vac' command on the fossil console would give
me a real, venti score...

Now that I understand where I may be wrong, the question is 'how to
initialize an empty fossil from a venti without knowing any scores?'.
My plans to have the nightly 'snap -a' score mailed offsite were supposed to
be implemented later. I just wasn't thinking I'd break things so early in the
game. Maybe that's the key -- I wasn't thinking :)

andrey

Re: [9fans] fossil/flfmt -v problem

[jmk]

On Thu Jul  3 23:37:21 EDT 2003, mirtchov@cpsc.ucalgary.ca wrote:
> ...
> Now that I understand where I may be wrong, the question is 'how to
> initialize an empty fossil from a venti without knowing any scores?'.
> ...

Look back in the 9fans archives around March, there was a discussion of
this I think and some scripts were posted to help if I remember correctly.

Re: [9fans] fossil/flfmt -v problem

[andrey mirtchovski]

On Thu, 3 Jul 2003 jmk@plan9.bell-labs.com wrote:

> Look back in the 9fans archives around March, there was a discussion of
> this I think and some scripts were posted to help if I remember correctly.

I saw the scripts (lastfossil and vacchain) however they don't help me much,
since 'fossil/flfmt -v' with a wrong score destroyed my fossil partition.

andrey

Re: [9fans] fossil/flfmt -v problem

[Fco.J.Ballesteros, nemo]

> I saw the scripts (lastfossil and vacchain) however they don't help me much,
> since 'fossil/flfmt -v' with a wrong score destroyed my fossil partition.

I have a cron that prints (daily) the vacchain. Is there a better way?
This is also what you're asking, if I understood correctly.

Re: [9fans] fossil/flfmt -v problem

[andrey mirtchovski]

that's what i was about to implement before i managed to reformat my fossil
partition. now that i know what's the best way to do it.... i wish i knew
it couple of days earlier.

i'm stuck with a fossil partition of which i know only a single vac score.
the more i learn about venti the more i realize that a venti without a
single score is pretty much useless -- one needs to know where the root
block is -- so, a warning to all: keep your venti scores.

andrey

On Fri, 4 Jul 2003, Fco.J.Ballesteros wrote:

> I have a cron that prints (daily) the vacchain. Is there a better way?
> This is also what you're asking, if I understood correctly.
>

Re: [9fans] fossil/flfmt -v problem

[jmk]

On Fri Jul  4 09:52:25 EDT 2003, mirtchov@cpsc.ucalgary.ca wrote:
> ...
> block is -- so, a warning to all: keep your venti scores.
> ...

a console with real paper in it makes an excellent log.

Re: [9fans] fossil/flfmt -v problem

[Fco.J.Ballesteros, nemo]

[-- Attachment #1: Type: text/plain, Size: 109 bytes --]

I use a printer. When I said a cron to print vac scores
I did actually mean "print" :-) (no joke, really).

[-- Attachment #2: Type: message/rfc822, Size: 1854 bytes --]

From: jmk@plan9.bell-labs.com
To: 9fans@cse.psu.edu
Subject: Re: [9fans] fossil/flfmt -v problem
Date: Fri, 4 Jul 2003 12:42:54 -0400
Message-ID: <cc60cbcd83a8e73e318697e8eb2707db@plan9.bell-labs.com>

On Fri Jul  4 09:52:25 EDT 2003, mirtchov@cpsc.ucalgary.ca wrote:
> ...
> block is -- so, a warning to all: keep your venti scores.
> ...

a console with real paper in it makes an excellent log.

Re: [9fans] fossil/flfmt -v problem

[rog]

> a console with real paper in it makes an excellent log.

but...  watch out, as a venti score is a dangerous thing.

given a venti archive score and access to a venti server, you've got
access to all data in the filesystem, regardless of permissions or
ownership.

i wonder whether it might be sensible to print venti scores blinded
with some private key, or something along that sort of line?

Re: [9fans] fossil/flfmt -v problem

[Fco.J.Ballesteros, nemo]

[-- Attachment #1: Type: text/plain, Size: 121 bytes --]

I print them on a printer sitting in a locked room. If they break the
room they already could break  the servers, sic.

[-- Attachment #2: Type: message/rfc822, Size: 2018 bytes --]

From: rog@vitanuova.com
To: 9fans@cse.psu.edu
Subject: Re: [9fans] fossil/flfmt -v problem
Date: Fri, 4 Jul 2003 18:21:44 +0100
Message-ID: <63090fce9e1be3b8bc05488eefe1e15f@vitanuova.com>

> a console with real paper in it makes an excellent log.

but...  watch out, as a venti score is a dangerous thing.

given a venti archive score and access to a venti server, you've got
access to all data in the filesystem, regardless of permissions or
ownership.

i wonder whether it might be sensible to print venti scores blinded
with some private key, or something along that sort of line?

Re: [9fans] fossil/flfmt -v problem

[Fco.J.Ballesteros, nemo]

[-- Attachment #1: Type: text/plain, Size: 256 bytes --]

Perhaps an alternative would be to keep the fossil vacs
logged in a separate block in the fossil (perhaps venti) disk and keep
that block as a circular log. That way we could avoid
printing the vacs; you could cat the log to recover a
fossil particion.

[-- Attachment #2: Type: message/rfc822, Size: 4140 bytes --]

[-- Attachment #2.1.1: Type: text/plain, Size: 121 bytes --]

I print them on a printer sitting in a locked room. If they break the
room they already could break  the servers, sic.

[-- Attachment #2.1.2: Type: message/rfc822, Size: 2018 bytes --]

From: rog@vitanuova.com
To: 9fans@cse.psu.edu
Subject: Re: [9fans] fossil/flfmt -v problem
Date: Fri, 4 Jul 2003 18:21:44 +0100
Message-ID: <63090fce9e1be3b8bc05488eefe1e15f@vitanuova.com>

> a console with real paper in it makes an excellent log.

but...  watch out, as a venti score is a dangerous thing.

given a venti archive score and access to a venti server, you've got
access to all data in the filesystem, regardless of permissions or
ownership.

i wonder whether it might be sensible to print venti scores blinded
with some private key, or something along that sort of line?

Re: [9fans] fossil/flfmt -v problem

[rog]

> I print them on a printer sitting in a locked room. If they break the
> room they already could break  the servers, sic.

it has to be said, though, that it's likely to be considerably easier
to scribble down 40 hex digits while the room is temporarily unlocked
than to break in to the server and steal all the data there and
then...