Redmine

Redmine will load the complete file in memory before sending it to the client, so your server needs to be able to allocate that much memory in addition to the memory rails and redmine already use. You could try serving your files through WEBDAV on apache, which you could integrate to the redmine authentication and authorization through the redmine.pm provided for svn integration (svn over http is just a superset of webdav, so it should work).

Thanks for the answer; excuse me if my question is too trivial, but I never worked with redmine.pm and webdav: I'm on a hosting, and I have no access to httpd.conf, is your solution applicable in this case?

Mmh, good question. I think you need stuff in the vhost.conf and mod_perl loaded. But even then, it's not "simple" so I'm not sure if it would be the right solution for you if you have no experience with webdav.

I asked to hosting's support if I have memory limitation on Mongrel, and they answer that I've no limit, so maybe the error in my first post came for other reasons. So, now I have no idea on what could be the cause, what kind of checks can I do? Very thanks!

Mmh, the limitation could come from a lot of places: available memory, number of other processes, OS limit on per-process memory available… Anyway, what is approximately the size at which this problem begins to occur?

Thanks again for the answer; I've done more tests, and I can't download files bigger than 8 MB. So, the problem is very frustrating: in our team we need to manage with Redmine files of this size.

For the available memory:

[~]# free -m
             total       used       free     shared    buffers     cached
  Mem:          7986       7375        611          0        327       3410
  -/+ buffers/cache:       3637       4348
  Swap:         8000          1       7999

It's a report of a specific instant, but there is always much more than 250 MB available.

For the number of other processes, I'm on a shared hosting, and with ps aux I see many processes, tens of them. But the situation (from the memory point of view) seems untroubled:

[~]# ps auxw --sort -rss
  USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
  mysql    31281  5.5  9.9 1232644 809872 ?      Sl   Apr29 1131:05 /usr/sbin/mysqld [...]
  flxxxx   17956  0.0  1.6 262132 135168 ?       S    Apr18  10:41 /usr/bin/ruby /usr/bin/mongrel_rails start -p 12xxx -d -e production -P log/mongrel.pid
  690       8145  0.0  1.6 256424 131788 ?       S    May05   1:31 /usr/bin/ruby /usr/bin/mongrel_rails start -p [...]
  532      25012  0.0  1.5 249648 123020 ?       S    May04   0:10 /usr/bin/ruby /usr/bin/mongrel_rails start -p [...]
  752      16568  0.0  1.4 244648 119896 ?       S    Apr18   0:15 /usr/bin/ruby /usr/bin/mongrel_rails start -d -p [...]
  root      9136  0.0  1.3 145228 113380 ?       Ss   Apr15   3:27 cfenvd
  864      14221  8.2  1.2 376292 98404 ?        SN   17:20   0:00 /usr/local/apache/bin/httpd -DSSL
  864      14200  6.8  1.2 376284 98400 ?        SN   17:20   0:00 /usr/local/apache/bin/httpd -DSSL
  864      14159  7.1  1.1 375484 97612 ?        SN   17:20   0:00 /usr/local/apache/bin/httpd -DSSL
  [...]

I use about 135000 KB (my mongrel process is the first one), that is 1.6% of memory. What can I do? Thanks!

Well, that's no filesize that should cause problems. Just out of the blue, is your max filesize setting maybe near 8MB? If yes, try upping that and see if it makes any difference. If that doesn't work either, head over to IRC and see if anyone can help (edavis10, khalsa and rchady come to mind that could have an idea) or/and file a new bug, though you will need to describe how to reproduce this error, i.e. try on a fresh install if you get the same errors.

Felix Schäfer wrote:

is your max filesize setting maybe near 8MB?

Do you refer to this setting? If yes, it was setted to 40 MB.

If yes, try upping that and see if it makes any difference.

I tried, but I saw exactly the same behaviour.

Anyway, the problem is only in one way, that is, I can't download the files: for example, I can upload files without problems (and much bigger than 8 MB), and the uploads are correct (in the files folder of Redmine I can see the files uploaded); but when I try to download them (for example as wiki attachments, but the same happens in other areas) I receive the error.

If that doesn't work either, head over to IRC and see if anyone can help (edavis10, khalsa and rchady come to mind that could have an idea) or/and file a new bug, though you will need to describe how to reproduce this error, i.e. try on a fresh install if you get the same errors.

Thanks for the hints, I'll try your suggestions.

I just met this error too with a 320MB file on a Redmine 0.8.6. the Server had 1GB RAM and got the "NoMemoryError (failed to allocate memory):"
Because it's on a VMware I added another GB of memory and I could download the file.
For testing I started a "top" on the machine during download and saw that the ruby process allocated the memory but did not give it back! So with the 4th download I got the NoMemoryError again! Only a restart of the appliance helped!

hope that this error is fixed on newer versions were we'll move in about a month.

Ronald Theile wrote:

For testing I started a "top" on the machine during download and saw that the ruby process allocated the memory but did not give it back! So with the 4th download I got the NoMemoryError again! Only a restart of the appliance helped!

That's a problem with the ruby GC poorly handling large files, and that's why there's a setting to limit file sizes in the redmine configuration. If you need to work with such big files, you're probably better off working with a webdav or svn directory.