[mb-devel] building qemu vm from scratch, problems with search server indexing

[Eldon Tyrell]

hi,

i've been trying to get a working qemu vm for about a week now, after
giving up on the available vbox 32bit vm, i've decided to build one from
scratch (debian 7.8 64bit).

i'm following your vagrant vm setup and if i can i would like to make
that vm available for your resources but i have one problem i can't fix.

following git INSTALL doc, musicbrainz-server install and db import work
fine, as far as i can tell the mb server works, haven't tested updates yet.

i managed to install the search server from the current bitbucket
INSTALL doc : http://www.bytebucket.org/metabrainz/search-server

installing works fine, with some minor debian apt sources tweaks, but
building indexes is problematic :

using the default documentation command fails :
java -Xmx512M -jar index-2.0-SNAPSHOT-jar-with-dependencies.jar
--indexes-dir /home/search/indexdata --freedb-dump /home/search/<freedb
tar.bz2 file>

i end up with a weird output as follows :

***@musicbrainz:~$ java -Xmx512M -jar
index-2.0-SNAPSHOT-jar-with-dependencies.jar --indexes-dir
/home/search/indexdata --freedb-dump
/home/search/freedb-complete-20150201.tar.bz2 --db-user musicbrainz
Index Builder Started:15:51:42
recording: Unable to get replication information
tmp_artistcredit:Started at:15:51:43
tmp_artistcredit:Finished:35.0 secs
tmp_artistcredit:Created Indexes:1.0 secs
tmp_release :Started at:15:52:19
tmp_release :Finished:96.0 secs
tmp_release :Created Indexes:2.0 secs
tmp_release_event :Started at:15:53:57
tmp_release_event :Finished:6.0 secs
tmp_release_event :Created Indexes:1.0 secs
tmp_track :Started at:15:54:05
tmp_track :Finished:3742.0 secs
tmp_track :Created Indexes79.0 secs
recording:Started at 16:57:50
recording:Isrcs Queries 114.0 secs6298 / 17426298 (100%)
recording:Track Queries 3903.0 secs
recording:Artists Queries 3216.0 secs
recording:Track Artists Queries 6144.0 secs
recording:Releases Queries 2740.0 secs
recording:Recording Queries 142.0 secs
recording:Build Index 238.0 secs
recording:Build Store 1138.0 secs

recording:Finished:26010.0 secs
recording:Started Optimization at 00:11:21
release:Started at 00:11:21
recording:14520237 db rows:14520237 lucene docs
recording:Finished Optimization:4.0 secs
Killede:Indexing 920000...939999 / 1565285 (60%)
***@musicbrainz:~$


The recording index dir is about 8GB big but other indexes are empty so
i guess it crashed. For example artist searches would give no results.
I tried to build each (not recording) index one by one and they all pass.

--test works

Trying to build "recording" fails every time but with very little output.
The vm has 2G ram and 120G disk space (i got one "no space left" with my
initial 80G size) but as far as i can tell ram is not a problem, it
barely uses over 600MB and never swaps.

i'm currently trying another run with the --debug option to see if i get
more output but if you could point me to some other options or logs, it
would help.

thank you,
eldon