Mediatex 0.9: Stack model

3.6 Meta-data consistency

GIT pull and push operations are done implicitly when loading and serialising files. This may be view like a stack model ensuring consistency on meta-data files (see Relational schema).

The so-called meta-data files are mentioned bellow using the followings numbers:

/etc/mediatex/mdtx.conf (see mdtx.conf)
~mdtx/git/mdtx/supports.txt (see supports.txt)
~mdtx/md5sums/mdtx-COLL.md5 (see mdtxCOLL.md5)
/etc/mediatex/mdtx-COLL/catalogXXX.txt (see catalog.txt)
/etc/mediatex/mdtx-COLL/extractXXX.txt (see extract.txt)
/etc/mediatex/mdtx-COLL/servers.txt (see servers.txt)

All calls to CLIENT follow the same operations:

Stage	Description
[A	Scripting, which do not need to load the meta-data
[B	GIT pull and parse configuration meta-data (1, 2)
[C	GIT pull and parse collection meta-data (3, 4, 5, 6)
[D	Generic upgrade works: upgrade SSH and GIT settings and re-compute the local image’s score
D]	Working on meta-data
C]	Serialise collection meta-data and GIT push (3, 4, 5, 6)
B]	Serialise configuration meta-data and GIT push (1, 2)
A]	Send `HUP` message to SERVER (see Server)

Notice:

Stage [D (Generic upgrade works) is only processed when (6) is loaded. This stage insure connection consistency between remote servers :

   MASTER          shared           SLAVES
---------------------------------------------
  mdtx.conf <--> servers.txt <--> mdtx.conf
                     |
.ssh/       <--------+----------> .ssh/
.git/config <--------+----------> .git/config

On stage C] (Serialise meta-data files), the default behaviour is to not serialise (4) and (5) unless they provide a new NN.txt post-fixed file. This intents to speed-up some queries endings (as make). In order to force serialising theses files you have to use the upgrade query.

Bellow details how meta-data files are affected by CLIENT’s queries.

ADMIN (see Who) queries:

Query	[A	[B	[C	[D	D]	C]	B]	A]
# mediatex adm init	y						1,2	y
# mediatex adm remove	y
# mediatex adm purge	y
# mediatex adm add user `user`	y
# mediatex adm del user `user`	y
# mediatex adm add coll [`serv`-]`coll`[@`host`[:`port`]]	y	1,2	6	y	y	6	1,2	y
# mediatex adm del coll `coll`	y	1			y		1	y

CLIENT to himself queries (or debugging queries):

Query	[A	[B	[C	[D	D]
$ mediatex adm update [coll `coll`]		1
$ mediatex adm commit [coll `coll`]		1
$ mediatex adm make [coll `coll`]		1,2	4,5,6	y	y
$ mediatex adm bind	y
$ mediatex adm unbind	y
$ mediatex adm mount `iso` on `path`	y
$ mediatex adm umount `path`	y
$ mediatex adm get `path` as `coll` on `path`		1	6	y	y

Queries to operate on SERVER:

Client query	[A	[B	[C	[D	D]	C]	B]	A]
$ mediatex srv save								y
$ mediatex srv extract								y
$ mediatex srv notify								y
$ mediatex srv [quick] scan								y
$ mediatex srv trim								y
$ mediatex srv clean								y
$ mediatex srv purge								y
$ mediatex srv status								y

PUBLISHER data management’s queries:

Client query	[B	[C	[D	D]	C]	B]	A]
$ mediatex add supp `supp` on `path`	1,2			y		2
$ mediatex add file `path`	1,2			y		2
$ mediatex del supp `supp`	1,2	6	y	y	6	1,2	y
$ mediatex ls supp	1,2			y
$ mediatex note supp `supp` as `text`	1,2		y	y		2
$ mediatex check supp `supp` on `file`	1,2	5		y		2
$ mediatex add supp `supp` to (all\|coll `coll`)	1,2	6	y	y	6	1,2	y
$ mediatex del supp `supp` from (all\|coll `coll`)	1,2	6	y	y	6	1,2	y
$ mediatex upload[+] [file `file` [as `target`]]* [catalog `file`] [rules `file`] to coll `coll`	1	5		y	5		y

PUBLISHER meta-data management’s queries:

Client query	[A	[B	[C	[D	D]	C]	B]	A]
$ mediatex ls [master] coll		1			y
$ mediatex motd		1,2	3,5,6	y	y	6	2	y
$ mediatex add key `file` to coll `coll`		1	6	y	y	6	2	y
$ mediatex del key `fingerprint` from coll `coll`		1	6	y	y	6	2	y
$ mediatex upgrade[+] [coll `coll`]		1	4,5,6	y		4,5,6	1,2	y
$ mediatex make [coll `coll`]		1	4,5,6	y	y	4,5,6	1,2	y
$ mediatex clean [coll `coll`]	y
$ mediatex su [coll `coll`]		1			y
$ mediatex audit for `mail` coll `coll`			5