Memory Problems in Editor
Brought to you by:
ddepinet,
rameshdanala
Menu
▾
▴
#70 Memory Problems in Editor
We currently have around a dozen developers
developing with jewelbox to create model and service
files. What I have noticed is that as the .md-xml files
(query files specifically) get larger than, say 3Mb - the
editor starts failing - it wont load them. Some of this is
able to be worked around by pumping up the -Xm args
in the startup scripts to obscenely large numbers and
shutting down everything but JewelBox, however - at
some point, even that appears to stop working reliably.
Has anyone else reported or looked into this? I would
include a file, but of course, it would not be an accurate
representation because our databases are unavailble to
the outside world.
Logged In: YES
user_id=718471
Another client as md-xml files > 4M and does not have this
problem, even without bumping -Xm args. They are on V
1.0.0, but I have no problems with V 1.2.1 editing and
saving their domain model. One key difference is that they
don't use the Query service; if you don't do any "Test" with
the query service do you still have the problem? (Perhaps
the Query service is using too much memory?)
Logged In: YES
user_id=1231164
Dave,
Yes - I have md-xml files that are >4M as well and I can run
pretty much out of the box as well. It's only when we have a
query.md-xml file for the query service that it seems to flop.
Is this related to the issue posted by you shortly after your
response to this one?
-Brian
Logged In: YES
user_id=718471
Brian,
Let me make sure I understand this:
- You've got this file (query.md-xml). You start up
ModelEditor (with or without -Xm args) and try to open the
file without doing anything else first and it fails to load.
Is that right?
- If that's the problem, then you could send me the file and
I could try to reproduce it-- I wouldn't need access to any
database because it doesn't try to access the database just
to load the file. If you're willing to do that, you could
zip it up and send to ddepinet@10xsoftware.com. If I can
reproduce this problem, then I can find out what's wrong and
address it either with a patch or at least in a future release.
Also, if you send the file, let me know what JDK version and
JewelBox version you're using.
(I don't think the JDBC batching issue is related to this--
that issue just affects overall query performance when
returning very large numbers of records.)
Thanks,
Dave
Logged In: YES
user_id=1231164
I have gotten to the bottom of this - sort of. It seems that this
is totally dependant on the number of table references or the
like currently in the model editor. In any kind of real
environment it seems to be a big mistake to just import all
the tables - Jewelbox seems to burn out even with memory
settings set very, very high.
Logged In: YES
user_id=1231164
I have gotten to the bottom of this - sort of. It seems that this
is totally dependant on the number of table references or the
like currently in the model editor. In any kind of real
environment it seems to be a big mistake to just import all
the tables - Jewelbox seems to burn out even with memory
settings set very, very high.
Logged In: YES
user_id=718471
OK, thanks for isolating this. Can you give me some idea of
how many tables and columns you have? Perhaps there's some
inefficiency in the way those are handled.
Logged In: YES
user_id=1231164
I'm really not sure at what point it tips the scale, sorry. But
with regard to file-size, most of our query files end up being
around 4-5 MB if we include the whole schema, but only a
couple hundred K if we selectively choose the tables we
need. Maybe I can try to grab some metrics for you if I can
get a moment free (unlikely).