Project

General

Profile

Connection Reset After upgrade

Added by Freddy Jones over 7 years ago

Hello all.

I just upgraded from 3.2 to 3.3. Due to ruby incompatibilities I had to upgrade on a separate box. Other than that I followed the upgrade directions - RedmineUpgrade

Other than redmine we don't use rails or ruby, so my experience only limited to redmine. We are a small environment hosted behind a firewall so I just run it under webrick, I know it needs to be hosted better, but all attempts so far have failed, and webrick works for us.

So far everything has been working with one exception, when trying to download a file attachment from an issue I get a "connection reset" in the browser. This happens to all files on all issues. I have checked the file is there and the redmine process has access to the files, according to the logs things look correct with the exception of 0.0ms to send a file:

Started GET "/attachments/download/437/abcd.csv" for 10.45.45.45 at 2017-06-20 18:38:51 +0000
Processing by AttachmentsController#download as HTML
  Parameters: {"id"=>"437", "filename"=>"abcd.csv"}
  Current user: freddy.jones (id=11)
Sent file /opt/redmine-3.3.3/files/2017/03/170329080336_abcd.csv (0.0ms)
Completed 200 OK in 20ms (ActiveRecord: 11.0ms)

I have been told that new files can be uploaded and both new and old files can be deleted, just not downloaded. I do have imagemagick and rmagick available, I am not sure it was available in the old version, if that makes a difference.

Anyone have any suggestions ?

Thanks
Freddy


Replies (1)

RE: Connection Reset After upgrade - Added by Freddy Jones over 7 years ago

Well, a few more details.

Our old instance was running on ruby 2.0 During install one of the gems said it wouldn't run on 2.0 so I upgraded to ruby 2.3, which caused the need to go to different hardware. It appears that we can download files after the webrick is restarted. Once it has been up and running for a while we start getting the connection resets. If I restart the service all starts working again. Maybe a memory leak ?

Anyone ??

    (1-1/1)