• User manual: http://www.indexdata.com/zebra/doc/
• Koha Indexing Configuration: http://koha.org/documentation/manual/3.2/searching/guide-to-searching/indexing-configuration

Zebra start i stop

mglavica@koha-upgrade:~$ sudo /usr/share/koha/bin/koha-zebra-ctl.sh start
Starting Zebra Server

mglavica@koha-upgrade:~$ sudo /usr/share/koha/bin/koha-zebra-ctl.sh stop
Stopping Zebra Server

Obnavljanje zapisa u indexu

Za to bi trebao služiti Zebraqueue Daemon, ali ne radi kako treba (koha-zebraqueue-ctl.sh). Umjesto toga index se rebuilda svako malo uz pomoć crona

Pokrenuti kao user koha.

KOHA_CONF=/etc/koha/koha-conf.xml
PERL5LIB=/usr/share/koha/lib

# m h  dom mon dow   command
*/4 * * * *   ps ax | grep -v grep | grep rebuild_zebra.pl || /srv/koha/misc/migration_tools/rebuild_zebra.pl -b -z >/dev/null

Dodavanje polja u pretraživanje

etc/zebradb/marc_defs/marc21/biblios/record.abs
mapping of MARC fields to indexes

etc/zebradb/biblios/etc/bib1.att
list of search indexes and their corresponding Z39.50 use attributes

etc/zebradb/ccl.properties
for searching purposes

UTF-8 to ASCII mappings

http://lists.indexdata.dk/pipermail/zebralist/2007-August/001707.html

Dijakritički znakovi

Situacija: u bazu su povučeni zapisi iz NSK (Voyager) u kojima su naša slova s dijaktiticima zabilježena kao dva znaka (slovo bez dijakritika + dijaktitik).

Nije lijepo da takvi znakovi završe u Koha bazi, ali kad su već unutra, treba ih nekako učiniti pretraživima.
U datoteku /etc/koha/zebradb/etc/word-phrase-utf.chr dodati ekvivalente:

git

# Characters to be considered equivalent for searching purposes
equivalent č(cˇ)
equivalent Č(Cˇ)
equivalent ć(c´)
equivalent Ć(C´)
equivalent š(sˇ)
equivalent Š(Sˇ)
equivalent ž(zˇ)
equivalent Ž(Zˇ)

Sad to dobro radi kad se pretražuje s "normalnim" znakovima. Ostaje problem kod browsanja preko liknova - strgani zapisi nađu samo strgane (pr. http://10.60.0.253/cgi-bin/koha/opac-detail.pl?biblionumber=241488, Pavao Pavličić).

Zebra register i shadow

Veličinu zebra register i shadow treba podesiti sukladno veličini baze. Za oko 120.000 zapisa, dovoljno je 4G. Za FF je sada podešeno 16GB.

Z39.50 server

Dokumentacija

• http://koha.org/documentation/manual/3.0/administration/additional-parameters/z39.50-servers
• http://lists.indexdata.dk/pipermail/zebralist/2009-March/002142.html
• http://bugs.rot13.org/rt/Ticket/Display.html?id=377

Ali sve je uzalud jer je pravi odgovor na http://www.indexdata.com/zebra/doc/zebrasrv.html

The first two servers, "server1" and "server2", can be reached by both listener addresses - since no listenref attribute is specified.

Pravo rješenje

dpavlin@koha-dev:~$ yaz-client 
Z> open localhost:9999/biblios
Connecting...OK.
Sent initrequest.
Connection accepted by v3 target.
ID     : 81
Name   : Zebra Information Server/GFS/YAZ
Version: 3.0.52 e687cb7eb87c841f0d1a374174d51d30371f2d97
Options: search present delSet triggerResourceCtrl scan sort extendedServices namedResultSets
Elapsed: 1.605266
Z> find human
Z> show 1
Z> quit
See you later, alligator.

dpavlin@koha-dev:~$ yaz-client -c /etc/koha/zebradb/ccl.properties localhost:9999/biblios
Connecting...OK.
Sent initrequest.
Connection accepted by v3 target.
ID     : 81
Name   : Zebra Information Server/GFS/YAZ
Version: 4.1.7 c43e561fdfe12989a6040163dae0e28ba0a02453
Options: search present delSet triggerResourceCtrl scan sort extendedServices namedResultSets
Elapsed: 0.001445
Z> querytype ccl2rpn
Z> find (kw,wrdl,rtrn=human ) not Suppress=1
Sent searchRequest.
Received SearchResponse.
Search was a success.
Number of hits: 1935, setno 1
SearchResult-1: term=    human cnt=1935, term=    1 cnt=21
records returned: 0
Elapsed: 0.025643
Z> 

# add all to http://www.loc.gov/z3950/agency/defns/bib1diag.html
ZEBRAOPTIONS="-v none,fatal,warn,all"

dpavlin@koha-dev:~$ sudo /etc/init.d/koha-zebra-ctl.sh restart
Restarting the Zebra Server

dpavlin@koha-dev:~$ sudo tail -f /var/log/koha/koha-zebradaemon-output.log | grep ERROR
18:45:58-29/03 zebrasrv(1) 605aecf5 [request] Init ERROR 1011 ID:81 Name:ZOOM-C/YAZ Version:4.1.7 c43e561fdfe12989a6040163dae0e28ba0a02453

Problemi

Neki zapisi nestaju u katalogu

Nepoznat obrazac u zapisima koji fale.

Nagađanje 1:

da li su problematični oni zapisi koji imaju različit biblionumber i biblioitemnumber?
To je dozvoljena kombinacija, ne bi trebalo smetati.
Problem je u nečem drugom, no ipak smo odlučili te zapise poindeksirati ponovo, ovako:

mysql> create temporary table foo as select max(id) as id from zebraqueue join biblioitems on biblioitems.biblionumber=zebraqueue.biblio_auth_number and biblionumber <> biblioitemnumber group by biblio_auth_number ;
mysql> update zebraqueue set done = 0 where id in (select id from foo) ;

Nakon toga pokrenuti rebuild_zebra.pl.

I dalje postoje zapisi koji se ne vide u katalogu.

Isprobano je ponovno indeksiranje zapisa i na jednom setu, ne znam kojem :(, zebra je javila ovo:

03:44:43-29/01 zebraidx(11983) [log] MARC: Bad directory
03:44:43-29/01 zebraidx(11983) [warn] MARC: Base address does not follow directory
03:44:43-29/01 zebraidx(11983) [warn] MARC: Bad offsets in data. Skipping rest
03:44:43-29/01 zebraidx(11983) [warn] Record didn't contain match fields in (bib1,Local-number)
03:44:43-29/01 zebraidx(11983) [log] error grs.marcxml.record /tmp/7EOcJqT94o/upd_biblio/exported_records 1580603

Nagađanje 2:

Potrgan MARC

• napraviti dump MARC zapisa iz Kohe (uz pomoć WebPAC2)
• provjeriti s marclint

Out of memory

• http://old.nabble.com/reindex_zebra.pl---out-of-memory-td19315345.html