Why I am not able to upload large files to confluence?

[Ruben Akopyan]

Dear community.

We have a problem on one of our confluecne installations. When we try to upload large files (100Mb and larger) the upload indicator hangs or restarts. Sometimes it gives error - Server is not available.

We have increased upload limit to 10G and put very large timeout but it did not help. There is no info in logs as well. Looks like logs are not even logged for this error. (if you check Logs file inside confluecne home directory).

Installation is on CentOS 6 + Postgres DB.

If someone could guide me how to solve the problem or at least find the source of the problem I would be very grateful.

Regards!

[Nic Brough (Adaptavist)]

Confluence is not a file server, it's not built for handling large files as attachments, and doesn't have much resilience. With a large limit and a large timeout, it can accept big files, but if there's any issue with the upload, even the slightest hitch, it doesn't handle it, and the transfer dies. The slightest wobble in your network connection can do this.

You should be looking at storing big files on something that is designed for it.

[jose.aleixo]

How us resolve this big problem, we need post setup up 1GB.

[Nic Brough -Adaptavist-]

>we need post setup up 1GB.

You almost certainly do not "need" to do this.  You need to share a file and are trying to do it the wrong way.  Files of that size have no practical use in Confluence, other than sharing, which is better done in something designed for sharing large files.

If you insist on doing this, then you will need to massively increase the memory available to Confluence and increase the timeout settings in its Tomcat.

[jose.aleixo]

Thank you so much Nic.

I believe that best solution is put package into file server "x", for example FTP Server and in Confluence one link, but I don´t will have a counter of the access, that´s right this?

[jose.aleixo]

How are the server license type packages distributed for Atlassian Jira, Confluence products?

[Nic Brough -Adaptavist-]

Yes, a service that is designed for large files is the answer.  You will in fact get a more accurate "count" from a file service because in Confluence, you can't tell if a file was successfully downloaded or if it were just viewed rather than downloaded.

I don't know what Atlassian use to serve up their installation files, but I would suspect it's something like Artifactory or Nexus.

[jose.aleixo]

Hello Nic, I understood and very grateful.
About package servers, I'll be researching in more detail, I really believe it's the Nexus.

Tks.