Indeed, the meta generation is taking too long, causing
the connection to MySQL to die in between trains.
Generating metadata for /home/babel-working/live/output/ganymede.
Array
(
[0] => Failed attempt to open database: babel - aborting
MySQL server has gone away
)
I have removed the meta data generation steps so that we can get
language packs and a basic update site working. That will buy us some
time to figure out the problem with the p2 stuff.
Antoine, thanks for your help,
Denis
Antoine Toulme wrote:
Well, it should be ant's log, since it's Ant here. I don't
know about that, sorry. It might have to do with the way we run ant in
generate1.php.
On Wed, Apr 22, 2009 at 21:32, Denis Roy <denis.roy@xxxxxxxxxxx>
wrote:
Antoine Toulme wrote:
For ganymede, the content.jar which is just
a jar of
content.xml weighted 65 Mbytes. So no, this is not normal I believe.
Anything in the log ?
The generate1.php log contains no useful info...
. . Completed language pack for Pseudo Translations (en_AA)
Generating metadata for /home/babel-working/live/output/galileo.
(usually freezes here)
. . Generating language pack for ganymede - Arabic (ar) (language_id=25)
etc...
Is there a log for Java and/or Eclipse?
On Wed, Apr 22, 2009 at 21:18, Denis Roy
<denis.roy@xxxxxxxxxxx>
wrote:
The
Galileo site.xml is 178KB.
If I watch the directory where artifacts.xml is being generated, I
notice that the file size of artifacts.xml varies throughout the
process. 1M, 500K, 212K, 712K, 0, 900K, 1.2M, 300K, and so on.
Constantly. Meanwhile, content.xml is steady at 473 bytes. Does this
sound like the correct behaviour to you?
Thanks for your input.
Denis
Antoine Toulme wrote:
You know, what the metadata generator
does,
using Ant, is
taking a site.xml and generating a content.xml file and an
artifacts.xml file, that is jars up.
So this might be a bit too much intense... can you check
the
size of the site.xml generated for Galileo ? I believe we could try to
upgrade the version of the p2 metadata generator in use, and if that
doesn't work, file a bug against p2.
On Wed, Apr 22, 2009 at 15:24, Denis
Roy <denis.roy@xxxxxxxxxxx>
wrote:
Team,
I am having great difficulty running a build from the live database.
My most recent attempt failed, as the process seemed to stall at this
stage for over 36 hours:
Generating metadata for /home/babel-working/live/output/galileo.
I'm not sure what the problem is since I don't know what is involved in
all the p2/metadata generation. Or perhaps it's normal for Generating
Metadata to take so long? Of course, the last time the live build has
run, we had fewer than 5% of the Galileo translations, whereas now we
have >90% for many languages.
In my troubleshooting process, I learned that if I LIMIT the number of
files to process to one or two files, or if I LIMIT the number of
languages in each train to one or two languages, the entire build
completes successfully for all the trains. I have since increased the
PHP memory limits from 64M to 512M, but the process stalls within
Java/Eclipse.
If anyone has any ideas, or can offer some help, it would be most
welcome. Until then, our available downloads are crippled, limited to
the unreleased Galileo, leaving all our Ganymede and Europa users in
the dark.
Denis
--
Denis Roy
|
