[Replicator-users] A look in "repli-update"

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Hello Jerome,

On Sat, Oct 26, 2002 at 09:04:28PM +0200, Jerome Warnier wrote:
> >At first I feel, that there is less concept or less (detailed)
> You do not mean "less", but "not much".

Ok, my english is not the best.

> I can help documenting the whole lot, but starting with repli-sync and=20
> repli-update is a good idea, as I'll have to use them soon also, and=20
> have recently put an eye on them.

Ok, lets start with repli-update.

> Sadly, I'm not someone who makes things by itself, I always need a reason.
> The maintainers are good-willing, off course, but have other things to=20
> do right now, as I understood.

Ok, I hope, that there is at leat a bit more of dokumentation
afterwards, to make it easier for other people to work on or use
replicator.

> >What do you think the next step should be?
> We should contact each other and discuss (on IRC) about the code (I know=
=20
> it enough to reply to common questions).

I have never used IRC and I'm not shure if I will like it. I like using
a dictionarie wenn writing english texts and this slow down my
typespeed very much.

I, think we should do the following things next. (No matter, how we
communicate). I already started a bit.

--------------------------------------------------------------------
1) Write a short overview of how "repli-update" should work.
--------------------------------------------------------------------

1. root-filesystem without @slash_exclude
It replicates the root (/) filesystem except for what is mentioned in
@slash_exclude. If you use cfengine, replicator parses
/etc/cfengine/cfengine.conf and does not copy files handled by cfengine
copying facility.=20

2. /usr-directory without @usr_exclude
It replicates the /usr directory (not the filesystem) except for what
is mentioned in @usr_exclude. Be aware that it may cross filesytems to
acheive this.

3. /var structure and files mentioned in @var_include=20
It replicates the directories structure of /var (only directories, no
files). It replicates the contents of the /var subdirectories mentioned
in @var_include.

This meschanism should be the same, no matter if repli-update is called
from repli-install or repli-sync

--------------------------------------------------------------------
2) Write a list of the Param "repli-update" understands and what they
should do.
--------------------------------------------------------------------

Because the first replication is working for me with repli-install but
not with repli-sync, and repli-sync uses other params to call
repli-update then repli-install, maybe there is a fault.

Correct me if I'm wrong, please.

--model <masterdir>
  Name of the Modell host, e.g. DNS-Name. No directory.

--destdir
  Directorie-tree which content should be synchroniced with the model.
  During repli-sync this should be "/"

--real
  Sets the variable $dryrun to "", else ist is $dryrun=3D"--dry-run"
  which is an option to rsync.

--dev
  Sets the variable $withdev=3D1 (true). This adds the string "/dev" to
  the list @slash_exclude bevor syncing the root-filesystem.

--nousr
  Sets the variable $nousr=3D1 (true). (Default is 0 ?????). This
  completly skips the replication of the /usr-directory.

--cfengine
  Sets variable $cfengine=3D1 (true). Bevor the sync of / the config from
  cfengine is copied from the model. Using the local cfengine all files
  handled by this tool are extracted and added to the list @slash_exclude

--noboot
  Sets $noboot=3D1 and skips a separately rsync for the /boot-directory.
  This rsync-run is necessary if /boot is as usual a own partition,
  because repli-update only syncs the root-filesystem, not the
  directory. (/boot is in the default @slash_exclude list to.)

-e
  Get the cfengine configuration from the model using ssh and scp
  instead of rsh and rcp.

--exclude
  Adds the argument to the @slash_exclude list.

--usr_exclude
  Adds the armunet to the @usr_exclude list.
 =20
--config
  Additional reads the given file. Can be used to set additional
  variables and override settings in /etc/replicator/replicator.conf,
  which is alway loaded first by repli-common.

repli-install calls:
repli-update --config $confile $ssh_opt --dev $rupdate_dry $netcopy_exclude=
 --model $model --destdir $installtarget
But I don't know, what is in the variables at that time.

repli-sync calls:
$share_dir_on_server/repli-update --destdir / --model $model --config $conf=
ile @ARGV
$confile is "/etc/replicator/repli-sync" on the replicated host and
$model ist taken from that file.

Why these calls behave different ?

--------------------------------------------------------------------
3) Take a deeper look in the code of "repli-update" and understand
what is doing what.
--------------------------------------------------------------------

rsync is startet with different params during repli-update. I don't
understand some things, maybe because of my perl knowledge. Can you
help me ?

dosystem("rsync $rsyncopt ${model}::replicator/boot/ $destdir/boot/. 2>&1");
Clear.

$excludefile =3D "/tmp/excl.rsync.$$";
$fh =3D new FileHandle $excludefile,"w" or die "opening exclude file:$!\n";;
foreach (@slash_exclude) { print $fh "- $_\n"; }
$fh->close; =20
dosystem("rsync $rsyncopt --one-file-system --exclude-from=3D$excludefile $=
{model}::replicator/ $destdir 2>&1");

The content of the list @slash_exclude is written to a file, line by
line. But for what is the "-" in { print $fh "- $_\n"; } ?

my $usr_exclude =3D '';
foreach (@usr_exclude) { $usr_exclude .=3D " --exclude '- $_' "; }    =20
dosystem("rsync $rsyncopt $usr_exclude  ${model}::replicator/usr/. $destdir=
/usr/. 2>&1");

Interesting in how many different way a list can be passed to rsync. Is
there a matter for the different ways? I have somewhere heard about a
maximum length of the command line. Do you know how long a command line
can be?

$dryrun or -d "$destdir/var/." or mkdir "$destdir/var",0755 or die "cannot =
make $destdir/var";;
dosystem("rsync $rsyncopt  --include '+ */' --exclude '- *' ${model}::repli=
cator/var/. $destdir/var/.");
foreach (@var_include) {
dosystem("rsync $rsyncopt ${model}::replicator/var/$_/ $destdir/var/$_ 2>&1=
");
}

For what is the "+" and the "-" sign in the [ei][xn]clude-rules? Is ist
perl-syntax or parsed by rsync?

What more?

--------------------------------------------------------------------
4) Correct "repli-update" and/or call it with other params
--------------------------------------------------------------------

- Do you have write access to the CVS-repository?

- What do you think about a @change_hostname list with files in which
  the name of the model should be replaced with the name of the target
  host. This can be implemented in repli-update or the example
  repli-postint.
 =20
- The config files replicator.conf and update_rules are not copied to
  the target during repli-install bacause of the default
  @shlash_exclude list. If we need them for repli.sync we should remove
  /etc/replicator from this list.
 =20
- More ideas ?
 =20
 =20
Enough for today.
What du you think about that?

Regards,
Bernd