We have a system that uses Jira, and Jira stores attachments to tickets in
/opt/jira/jiraattachments. Under that directory is the project name
RRT, and under that is a directory of tickets. Thus, ticket
RRT1234 has its attachments under:
We have a monitoring system that fires off an alert when there are more than 30,000 items in the
/opt/jira/jiraattachments/RRT directory. Considering we have 900,000 Jira tickets, this isn't really a surprise.
On a programming level, I don't really see an issue. Jira doesn't open the entire directory and keep all those directories open. In fact, the structure is arranged, so that Jira can immediately find the directory that contains the attachments.
However, on an OS level, is there a problem with a single directory containing over 32K files? I can see problems writing shell scripts and attempting to parse this many files. I can see problems with
ls trying to read and sort all of those files. I know back in the MS-DOS 2.x days, a directory couldn't have more than 512 entries. But we're no longer in the disco era. I can't see an OS stumbling on something like this.
$ uname -r 2.6.18-238.el5 $ df -kT . Filesystem Type 1K-blocks Used Available Use% Mounted on 10.10.136.125:/vol/jira_prod nfs 83886080 58621352 25264728 70% /jira_prod